design of linear and nonlinear control state …
TRANSCRIPT
DESIGN OF LINEAR AND NONLINEAR CONTROL SYSTEMS VIA STATE VARIABLE FEEDBACK, W I T H APPLICATIONS
I N NUCLEAR REACTOR CONIlROL i
John W. Herring, Jr. Donald G. Schultz Lynn E, Weaver Robert E , Vanasse
ENGINEERING EXPERIMENT STATION COLLEGE OF ENGINEERING
THE UNIVERSITY OF ARIZONA TUCSON, ARIZONA
I I . I I I I I I I I I I I I I I I 1
b I
DESIGN OF LINEAR AND NONLINEAR CONTROL SYSTEMS V I A STATE VARIABLE FEEDBACK, WITH APPLICATIONS
I N NUCLEAR REACTOR CONTROL
Prepared Under Grant NsG-490 Nat ional Aeronautics and Space Administration
John W. Herring, Jr. Donald G. Schul tz Lynn E. Weaver Robert E. Vanasse
Nuclear Engineering Department The Universi ty of Arizona
Tucson, Arizona Feb rua ry 19 6 7
Engineering Experiment S t a t i o n The Universi ty of Arizona College of Engineering
Tucson, Arizona
I I I I I I I I I I I I I I I I I I I
Chapter
1
2
3
4
TABLE OF CONTENTS
Page
PREFACE ............................................... i v ABSTRACT .............................................. V
INTRODUCTION TO THE PROBLEM ........................... 1
In t roduc t ion .......................................... 1 H i s t o r i c a l Background ................................. 2 Approach t o the Problem ............................... 5 Organization .......................................... 7
DEFINITIONS AND STABILITY CRITERIA .................... 9
In t roduc t ion .......................................... 9
Def in i t i ons ........................................... 11 1 2
System Representation ................................. 9
The S t a b i l i t y C r i t e r i o n of Popov ...................... CLOSED LOOP DESIGN OF LINEAR SYSTEMS V I A STATE VARIABLE FEEDBACK ..................................... 18
In t roduc t ion .......................................... 18 The SVF Method--Matrix Formulation .................... 24 Applicat ions i n Nuclear Reactor Control S v s t e m Design.. 36 Gain I n s e n s i t i v e Systems .............................. 58 Applicat ions of Gain I n s e n s i t i v e Design t o Reactor Control ............................................... 62 Procedure When A l l S t a t e Variables Cannot B e Fed Back.. 69 Summary ............................................... 71
DESIGN OF NONLINEAR AND/OR TIME-VARYING CONTROL SYSTEMS V I A STATE VARIABLE FEEDBACK ........................... 73
In t roduc t ion .......................................... 73 The SVF Method for Nonlinear Systems .................. 74 Absolute S t a b i l i t y of t h e Resul t ing Systems ........... 78 Closed Loop Response of t h e Resul t ing System .......... 78 The SVF Method f o r Time-Varying Systems ............... 9 5 Summary ............................................... 96
ii
TABLE OF CONTENTS (Continued)
Chapter
5
6
Page
PRACTICAL LIMITATIONS AND EXTENSIONS OF THE DESIGN PROCEDURE ............................................ 98
In t roduc t ion ......................................... 98 S t r u c t u r a l S t a b i l i t y of t h e System ................... 99 E f f e c t of t h e Location of t h e Nonlinear and/or Time-Varying Gain .................................... 112
117 Design f o r F i n i t e Sec tors of S t a b i l i t y surmnary .............................................. 122
...............
CONCLUSIONS AND SUGGESTIONS FOR FURTHER WORK ......... 1 2 3
Conclusions .......................................... 123 125 Suggestions f o r Further Work .........................
LIST OF REFERENCES ................................... 127
iii
I i - 1 I I 1 I I I I I I '1 I I I I I I
PREFACE
This r e p o r t r ep resen t s t he completion of one phase of t he s tudy of
c o n t r o l system design, a study sponsored by t h e Nat ional Aeronautics and
Space Administration under Grant NsG-490 on research i n and a p p l i c a t i o n
of modem automatic c o n t r o l theory t o nuc lea r rocke t dynamics and c o n t r o l .
The r e p o r t i s intended t o be a self-contained u n i t and t h e r e f o r e r epea t s
some of the work presented i n previous s t a t u s r e p o r t s .
Po r t ions of t h e work were submitted t o t h e Department of E l e c t r i c a l
Engineering a t the Universi ty of Arizona i n p a r t i a l f u l f i l l m e n t of t h e
requirements f o r t h e degree of Doctor of Philosophy.
i v
I I . 1 I 1 1 I I I I I I I I 1 I I I 1
ABSTRACT
A method f o r designing l i n e a r closed loop c o n t r o l systems, and non-
l i n e a r c losed loop c o n t r o l systems having a s i n g l e , memoryless, non l inea r
and/or time-varying gain wi th an input-output c h a r a c t e r i s t i c confined t o
the f i r s t and t h i r d quadrants i s developed. Nonlinear systems designed
by t h e proposed method have the p rope r t i e s of a b s o l u t e s t a b i l i t y and
bounded ou tpu t s f o r bounded inpu t s f o r any non l inea r gain of t h e type
considered. Systems i n which t h e gain i s time-varying a l s o have t h e s e
p r o p e r t i e s i f t h e time-varying gain i s constrained t o a f i n i t e s e c t o r i n
the f i r s t and t h i r d quadrants. The l i n e a r p a r t of t he open loop system
can have no more than one i n t e g r a t i o n , and a l l o t h e r p o l e s must have
nega t ive real p a r t s . I n i t s bas i c form, t h e design method r e q u i r e s t h a t
t he non l inea r and/or time-varying gain be loca ted a t t h e inpu t end o f t h e
plsrit to 52 zont rc l lcd .
The design procedure is based on feeding back a l l t h e s t a t e v a r i a b l e s
through constant l inear ga in elements. An equ iva len t feedback t r a n s f e r
func t ion , H
of t he l i n e a r p a r t of t h e open loop system) l i n e a r a l g e b r a i c equat ions
which can be solved f o r the feedback c o e f f i c i e n t s . H (s) has n-1 zeros
which are forced t o be equal t o n-1 of t he poles of t h e l i n e a r p a r t of t h e
( s ) , from t h e output i s used t o determine n (n i s t h e o rde r eq
eq
forward t r a n s f e r func t ion , G(s). It i s shown t h a t G(s)H ( s ) has one po le
and no t more than one zero.
eq
Thus the Popov frequency c r i t e r i o n f o r abso lu t e
s t a b i l i t y i s s a t i s f i e d f o r a l l gains of t h e type considered.
V
i I I I I I 1
I I I
I
1
An equiva len t system f o r input-output r e l a t i o n s which cons i s t of a
f i r s t o rde r nonl inear and/or time-varying p a r t i n series with an n-1 c r d e r
s t a b l e , l i n e a r , t ime-invariant p a r t is der ived. This equiva len t system
is used t o determine t h e closed loop response f o r a known inpu t t o a system
designed by the proposed method. It is a l s o used t o show t h a t the closed
loop system has n-1 poles equal t o n-1 poles of t h e l i n e a r p a r t of t he open
loop system. This sugges ts another key s t e p i n t h e design procedure, t h a t
of fo rc ing n-1 of t he open loop poles t o be equal t o n-1 des i r ed closed
loop poles . Closed loop poles on o r nea r t h e imaginary axis are no t pe r -
mi t ted due t o s t r u c t u r a l s t a b i l i t y problems.
Problems chich might be encountered i n applying t h e design procedure
t o phys i ca l systems are discussed, inc luding app l i ca t ions t o mul t i reg ion
nuc lea r r e a c t o r c o n t r o l and t h e s t r u c t u r a l s t a b i l i t y problem mentioned
above.
when a l l t h e s t a t e v a r i a b l e s cannot be fed back, o r when the n o n l i n e a r i t y
is n o t i n the proper l o c a t i o n f o r t h e b a s i c procedure t o apply, are d i s -
cussed.
Modif icat ions i n the bas i c design procedure which might be used
The Popov theorem f o r absolu te s t a b i l i t y and t h e design of l i nea r
systems for a des i r ed closed loop response by feeding back a l l t h e s t a t e
v a r i a b l e s are included as background ma te r i a l . The mat r ix formulat ion i s
developed f o r designing l i n e a r systems, and t h e procedure i s extended t o
the design of l i n e a r ga in i n s e n s i t i v e systems. The procedure f o r designing
l i n e a r ga in i n s e n s i t i v e systems i s then extended t o nonl inear and/or t i m e -
varying sys t e m s . The proposed design procedure f o r non l inea r and/or time-varying
s y s t e m s i s app l i cab le t o svstems of any order . Examples are given t o
v i
I 1 I I I I I I I I I I 1 I I I 1 1 I
i l l u s t r a t e the design and ana lys i s procedures. The r e su l t s of an analog
computer s imula t ion of a system designed by t h e proposed method are given.
.
vi i
I I - 1 i I 1 I I I I I I I I I I I I I
CHAPTER I
INTRODUCTION TO THE PROBLEM
Introduction
The performance required of control systems and devices by modem
technology has resulted in the development of complicated systems and
devices which are not amenable to analysis and synthesis by the classical
linear techniques. In addition to the undesired nonlinearities which
arise in these systems, nonlinearities are often introduced purposely in
order to realize the desired performance better or more economically. No
general methods for the analysis and synthesis of such nonlinear systems
exist, Rather, the methods in use today can be applied only to particular
classes of systems or t o particular applications. This is not surprising
because of the wide variety of nonlinear systems.
T I ALLLrea~ed interest is also being manifested in systems with time- . varying parameters. One obvious reason f o r this interest stems from the
space program where certain parameters of the control system may vary over
very wide ranges as the air density and temperature through which a vehicle
is moving changes rapidly. If a parameter variation is dependent on the
input to the system, the system is nonlinear. If the variation is caused
by some effect other than the input, and if this effect can be expressed
independently of the input, the system is time-varying. The comment on
the lack of general analysis and synthesis techniques for nonlinear systems
applies to time-varying systems as well.
1
2
A d i sp ropor t iona te amount of c o n t r o l theory has always been
d i r e c t e d towards s t a b i l i t y ana lys i s . This i s p a r t i c u l a r l y t r u e of non-
l i n e a r c o n t r o l theory. The con t ro l engineer i s handicapped by t h e l a c k
of s y n t h e s i s procedures which assure n o t only the s t a b i l i t y of t h e r e s u l t -
i n g system, bu t o t h e r d e s i r a b l e ope ra t ing c h a r a c t e r i s t i c s as w e l l . The
development of a proposed syn thes i s procedure which can be used t o design
c e r t a i n non l inea r and/or time-varying systems i s t h e s u b j e c t of t h i s
d i s s e r t a t i o n .
H i s t o r i c a l Background
The p r a c t i c e s and techniques of c lass ical c o n t r o l theory involve
p r imar i ly frequency domain methods. The t r end i n modern c o n t r o l theory
has been away from t h e s e frequency domain methods t o a n a l y s i s and design
procedures based on t i m e domain methods. S ta te space concepts and tech-
niques involving ma t r ix equations are becoming i n c r e a s i n g l y important.
Kalman (1964) has used such techniques t o show t h a t a l i n e a r system sub-
j e c t t o a q u a d r a t i c performance index can be made optimum by feeding back
a l l t h e state v a r i a b l e s through constant c o e f f i c i e n t s . This r e s u l t has
been a key influence i n the proposals by Morgan (1963, 1966) and Schultz
(1966) t h a t l i n e a r systems b e designed f o r a des i r ed closed loop response
by feeding back a l l t h e s t a t e v a r i a b l e s i n t h e proper l i n e a r combination.
Other recent developments i n c o n t r o l theory involve a n a l y s i s and
design procedures f o r nonl inear systems. P r i o r t o t h e p a s t f i f t e e n t o
twenty yea r s , p r a c t i c a l l y a l l the l i t e r a t u r e on the a n a l y s i s and s y n t h e s i s
of feedback c o n t r o l systems was confined t o t h e cons ide ra t ion of l i n e a r
systems. Valuable techniques have been developed and made a v a i l a b l e f o r
~-
I I . I I I 1 I I I 1 I I 4 I I I I 1 1
3
t h e s tudy of l i n e a r sys t em c h a r a c t e r i s t i c s . I n r ecen t years, an inc reas -
i n g awareness on t h e p a r t of con t ro l engineers t h a t r e s t r i c t i o n of think-
i n g t o l i n e a r s y s t e m s imposes unnecessary l i m i t a t i o n s on t h e design of
c o n t r o l systems has r e s u l t e d in a concerted e f f o r t t o develop correspond-
i n g a n a l y t i c a l techniques f o r non l inea r systems.
theory f o r nonl inear systems, which behave d i f f e r e n t l y f o r d i f f e r e n t
i n p u t s , appears t o b e v i r t u a l l y impossible a t t h e p re sen t t i m e because of
t h e mathematical d i f f i c u l t i e s involved.
made i n t h e t h e o r e t i c a l aspects of t h e problem, most of t h e procedures
which look good i n theory become unwieldy when app l i ed t o p r a c t i c a l systems
of o rde r h ighe r than f i r s t o r second.
have been developed apply only t o c e r t a i n r e s t r i c t e d s i t u a t i o n s .
t h e more widely known methods i n use today are:
of n o n l i n e a r i t i e s about some operat ing po in t and t h e a p p l i c a t i o n of l i n e a r
theory t o t h e r e s u l t i n g system. (2) Graphical methods. (3) Numerical
methods. (4) Describing funct ion a n a l y s i s . ( 5 ) Computer methods, both
analog and d i g i t a l . ( 6 ) Second method of Liapunov. (7) Popov theory.
(8)
and dynamic programming.
A completely gene ra l
Although some progress has been
Also, most of t h e techniques t h a t
Some of
(1) The l i n e a r i z a t i o n
Optimization techniques based on t h e maximum p r i n c i p l e of Pontryagin
Evidence of t h e increased i n t e r e s t i n time-varying systems is
apparent i n t h e published r e s u l t s of Rozenvasser (1963), Bongiorno(1963),
Sandberg (1964), and Narendra and Goldwyn (1964). Higgins (1966) shows
how t h e work of t h e las t t h r e e a r e r e l a t e d t o t h e Popov c r i t e r i o n .
The Popov theory is of primary i n t e r e s t h e r e , with t h e i s o c l i n e
method and Liapunov'ssecond method a l s o being used. The Liapunov method
I
I 4
i l l u s t r a t e s t h e t r end of modem c o n t r o l theory t o t i m e domain a n a l y s i s
and design.
r i o n , b u t i t is derived from t i m e domain equations.
between Popov's r e s u l t s and the second method of Liapunov has been
e s t a b l i s h e d by Yakubovich (1962) and Kalman (1963fd.
The Popov s t a b i l i t y c r i t e r i o n is a freq3ency domain crite-
The r e l a t i o n s h i p
The i s o c l i n e method is a b a s i c g r a p h i c a l method which a p p l i e s
d i r e c t l y t o a f i r s t o r d e r equation of t h e form K - f ( x , t ) , It is
a p p l i c a b l e t o second o rde r equat ions of t h e form 2 + f(x,x) = 0. It is
known as t h e phase p l ane method when used with second o r d e r systems and
is discussed i n most books deal ing wi th non l inea r systems, i nc lud ing
T r u d (1955) and Gibson (1963).
Liapunov developed his Second Method i n t h e l a te n ine teen th
century, bu t i t w a s no t u n t i l the 1940's i n Russia and t h e e a r l y 1960's
i n t h i s country t h a t c o n t r o l engineers became i n t e r e s t e d i n t h e theory.
Standard Engl ish language references are t h e books by Hahn (1963) and
LaSa l l e and Lefshetz (1961), and t h e paper by Kalman and Bertram (1960).
The method has evoked widespread i n t e r e s t because of i t s gene ra l nature .
However, because of t h e d i f f i c u l t y i n f i n d i n g Liapunov func t ions
( e s p e c i a l l y t h e b e s t one), i t has no t been p o s s i b l e t o apply i t gene ra l ly
t o systems of o rde r h ighe r than two o r t h ree .
methods f o r gene ra t ing Liapunov func t ions are found i n Letov (1961),
Schultz and Gibson (1962), Marsolis and Vogt (1963), and Gibson (1963).
Some of t h e b e t t e r known
The theory of Popov (1961) involves a frequency domain s t a b f l i t y
c r i t e r i o n which has a simple graphical i n t e r p r e t a t i o n ,
is no t l i m i t e d by t h e o rde r of the system. It provides abso lu te s t a b i l i t y
information and t h e r e f o r e cannot b e used t o determine regions of s t a b i l i t y
I ts use fu lness
I I I I I I I I I I I I I I I 1 I I I
5
as i n t h e Second Method.
Aizerman and Gantmacher (1964) and Lefshetz (1965). Extensions t o t h e
Popov theory involving condi t ions on t h e s l o p e of t h e n o n l i n e a r i t y have
been made by Yakubovich (1965) , Brocket t and W i l l e m s (1965a,, 1969b) , Dewey and Ju ry (1966), and Dewey (1966). The i n t e r p r e t a t i o n of t h e
cr i ter ia r epor t ed i n t h e s e papers is much more d i f f i c u l t than t h a t f o r
n o n l i n e a r i t i e s w i th no r e s t r i c t i o n on t h e s lope . The r e s u l t s of Dewey
appear t o be i n t h e most useful form, whi le those of Brockett and W i l l e m s
are probably t h e most general .
Other r e fe rences on t h e Popov theory are
Approach t o t h e Problem
I n t h i u d i s s e r t a t i o n , some of t h e r e c e n t l y developed methods of
s t a b i l i t y a n a l y s i s f o r nonl inear and/or time-varying systems are combined
w i t h t h e modem c o n t r o l concept of feedingback a l l t h e s ta te v a r i a b l e s t o
develop a proposed design procedure f o r a c e r t a i n class of non l inea r and/
o r time-varying s y s t m . A method f o r determining t h e closed loop response
of t h e r e s u l t i n g system is alsc developed.
c a b l e t o s i n g l e inpu t , s i n g l e output , systems con ta in ing a s i n g l e memory-
la66 nonlinear and/or time-varying ga in whose input-output c h a r a c t e r i s t i c
is confined t o t h e f i r s t and t h i r d quadrants. A s i g n i f i c a n t f a c t o r of
both t h e design and a n a l y s i s procedure i s t h a t they are a p p l i c a b l e , i n a
p r a c t i c a l sense, t o systems of any order .
The des ign procedure is appl i -
The design procedure stems from t h e method of designing l i n e a r
systems f o r a des i r ed closed loop response as developed by Schultz (1966).
It makes use of both t h e i d e a of feeding back a l l t h e s t a t e v a r i a b l e s as
suggested by modem c o n t r o l theory and of series compensation as used i n
6
c la s s i ca l design procedures. The method of designing l i n e a r systems €or
a d e s i r e d closed loop response is f i r s t ex t ended . to a procedure f o r de-
s ign ing l i n e a r gain i n s e n s i t i v e systems. This procedure i s then used
t o develop a design procedure fo r the c iass u i iiol-iIiiieiii- md/Por t i m e
varying systems given above. The Popov s t a b i l i t y c r i t e r i o n i s used as
a s y n t h e s i s t o o l i n t h a t t h e l i n e a r part of t h e system i s compensated
by feeding back a l l t h e s t a t e v a r i a b l e s i n a l i n e a r comblnation such
t h a t t h e Popov s t a b i l i t y condi t ions -are always s a t i s f i e d with no r e s t r i c -
t i o n s on t h e non l inea r and/or time-varying Rain. Thus t h e abso lu te
s t a b i l i t y s e c t o r of t h e r e s u l t i n g system inc ludes t h e entTre f i r s t and
t h i r d quadrants. An equivalent system is developed which makes i t
p o s s i b l e t o 1) show t h a t t h e closed loop system hasabounded output f o r
bounded i n p u t s and 2) determine t h e output of a p a r t i c u l a r c losed loop
system f o r a known i n p u t , r ega rd le s s of t h e o rde r of t h e system. A l -
though the output of t h e closed loop system depends unon t h e p a r t i c u l a r
non i inea r and/or time-varysng gain chaiacteriatics, thc system is linear-
i z e d t o t h e e x t e n t t h a t (n-1) of t h e closed loop Doles remain f i x e d ,
independent of t h e gain. This r e s u l t s i n considerably more c o n t r o l over
t h e n a t u r e of t h e closed loop response than i s p o s s i b l e i n the usual non-
l i n e a r and/or time-varvinE system. n r e f e r s t o t h e o r d e r of t h e I f n e a r
p a r t of the ep~tem, as discussed i n Chapter 2 .
The equ iva len t system mentioned above c o n s i s t s of a f i rs t o rde r
non l inea r and/or time-varying par t and an ( n - 1 ) s t o rde r t ime- inva r i an t
l i n e a r par t .
t h i s equivalent system t o show t h a t a system designed by the proposed
method has bounded s o l u t i o n s , o r i s Lagrange s t a b l e , f o r bounded i n p u t s .
The Second Method of Liapunov i s u s e d i n conjunct ion w i t h
7
The isocline method is used to determine the output of the nonlinear and/
or time-varying gain from the equivalent system for a given inpat.
this information, the expression for any other variable in the svstem can
be found by linear transform methods.
With
Organization
The second chapter discusses the type of nonlinear svstem to he
considered. The stability of such systems is discussed and Popov's
stability criterion is given. A l s o , definitions of terms used in later
chapters are given.
Chapter 3 is devoted to linear svstems, and in particular to the
development of a design procedure for aaln insensitive systems. This
chapter provides the background for Chapter 4 where the design procedure
for gain insensitive systems is extended to nonlinear and/or time-varying
systems. A matrix formulation for the state variable feedback method of
design is developed and examples are included to illustrate the desiqn
_I_ procedures.
The basic design procedure of this dissertation is presented in
The procedure for designing gain insensitive systems is ex- Chapter 4 .
tended to systems containing nonlinear and/or time-varying gains.
stability properties of the resulting system are discussed, and an equiva-
lent system is developed which makes it possible to determine input-output
relations of the closed loop system. Examples are included to illustrate
the design procedure and the determination-of the closed loop response.
Chapter 5 discusses the structural stability of the systems de-
signed by the procedure of Chapter 4. That is, it considers the effect
The
8
on t h e system’s q u a l i t a t i v e behavior of s m a l l changes i n t h e parameters.
The p o s s i b i l i t y of ob ta in ing more restrictive r e s u l t s when those of
Chapter 4 cannot be achieved is discussed.
i l l u s t r a t e t h e proposed modified design procedures.
Examples are included t o
Chapter 6 conta ins conclusions and suRRestions f o r f u r t h e r re-
search .
CHAPTER TI
DEFINITIONS AND STABILITY CRITERIA
In t roduc t ion
The purpose of t h i s chanter i s t o p re sen t c e r t a i n s t a b i l i t v
c r i te r ia and d e f i n i t i o n s of terms which are used i n t h e following
chapters . The Popov s t a b i l i t y c r i t e r i o n is of p a r t i c u l a r i n t e r e s t as
i t is used both i n t h e design procedure and i n t h e a n a l y s i s of t h e re-
s u l t i n g systems. The c l a s s of systems t o which t h e c r i t e r i o n a n n l i e s
and which i s considered i n t h i s d i s s e r t a t i o n i s discussed. The extension
of t h e Popov theory dea l ing with time-varying systems and with c o n s t r a i n t s
on t h e s lope of t h e n o n l i n e a r i t v are a l s o given.
System Representat ion
The system configurat ion has t h e form shown i n Figure 2-1. N - f ( a ) i s a non l inea r ga in , G ( s ) i s t h e t r a n s f e r func t ion of a l i n e a r
system whose po le s are on t h e imaginary a x i s o r i n t h e l e f t h a l f p l ane , CI
(J i s t h e input t o t h e nonl inear ga in and u = f ( a ) i s t h e ou tpu t . u . i s
a l s o t h e input c o n t r o l f o r t h e l i n e a r system. Using t h e terminologv of
Aizerman and Gantmacher (1964) , t h e p r i n c i p a l ca se is t h a t case i n which
a l l t h e po le s of G ( s ) a r e i n t h e l e f t h a l f plane. The D a r t i c u l a r cases
are those cases i n which some of t h e po le s of G(s) a r e on t h e imaginary
a x i s and t h e o t h e r s are i n t h e l e f t h a l f plane. The s implest p a r t i c u l a r
case i s t h a t case i n which G(s) has a s i n g l e po le a t t h e o r i g i n and t h e
9
Figure 2-1. nth Order System w i t h One Nonlinearity.
10
1 I I I I I I I 1 I I I 1 1 I I 1 I I
11
other poles are in the left half plane. The order of G ( s ) is designated
by the letter n.
The system equations are written in the form
$ = Ax+bu (3-1) - ” ..
u = f(u)
u = - I. 5 T .,
where 5 is an n-dimentional state vector, b is an n-dimensional control - vector, A is the n x n matrix of the linear system G ( s ) , uand u are
the scalar input and output of the nonlinearity, and k is an n-dimensional
feedback vector.
..
The nonlinearities under consideration satisfy the conditions
for the principal case, or
O < < K u -
( 2 - 4 )
(2-5)
for the particular cases, and
f(0) = 0 . (2 -6 )
Definitions
Several definitions of stability and other terms used in this
dissertation are given in this section. The origin is assumed to be the
equilibrium point in the definitions of asymptotic and absolute stability.
Definition 2 - 1 : The origin i s globally asymptotically stable if, for any initial conditions, the system state always returns to the origin as t* m . Definition 2-2: For a given K, the class of systems defined by Equations 2-1 to 2 - 3 is said to be absolutely stable if for any system in this class, that is, for any f(a) which satisfies Equation 2-4 for the principal case or Equation 2-5 for the particular cases, the origin is globally asymptotically stable.
Definition 2-3 : The boundedness of all solutions is described as stability in the sense of Lagrange, or Lagrange stability.
1 2
The fol lowing c r i t e r i o n f o r Lagrange s t a b i l i t y based on the Second Method
of Liapunov is given bv Lasa l l e and Lefschetz (1961):
Theorem 2-1: L e t R be a hounded s e t containing the o r i g i n and l e t V( ) be def ined throyghout t h e complement of R . as 111~71 -+ + 00 and i f V < 0 throuchout t he complement of R , then t h e system 2 - = X(x) - - is Lagrange s t a b l e .
D e f i n i t i o n 2-4: S t r u c t u r a l s t a b i l i t p i s t h e propertv of a physi- c a l system such t h a t t h e qual i ta t ive na tu re of i t s one ra t ion remains unchanged i f parameters of t h e svstem are s u b j e c t t o small v a r i a t i o n s .
I f V + + Q)
D e f i n i t i o n 2-5: A system with open loop po le s on the imaginarv a x i s i s stable-in-the-limit i f i t is s t a b l e f o r t h e l i n e a r eain f ( o ) = ea, where E i s a r b i t r a r i l y s m a l l . C)n t h e s-plane, t h i s means t h a t t h e imaninary a x i s poles move i n t o t h e l e f t h a l f plane f o r a r b i t r a r i l y s m a l l l i n e a r gains i n the closed loop system.
D e f i n i t i o n 2-6: A p l a n t i s s a i d t o be completely c o n t r o l l a b l e i f f o r any to each i n i t i a l s t a t e r(t ) can be t r a n s f e r r e d t o any f i n a l s ta te x( t , ) i n a f i n i t e t i m % .
D e f i n i t i o n 2-7: An unforced p l an t i s s a i d t o be completelv observable on [ t , t f ] i f f o r given to and t every s t a t e x( to) can be d termine2 from t h e knowledRe of v ( t f on I t o , t f ' l , where y ( t ) = c x ( t ) i s the output. 9
- 5
D e f i n i t i o n 2-4 i s Riven by Cunningham (1958) and D e f i n i t i o n s 2-6
and 2-7 by Kreindler and Sarachik (1964). T t i s d e s i r a b l e t o express t h e
cond i t ions f o r a system t o be c o n t r o l l a b l e and observable i n terms of t h e
c o e f f i c i e n t s i n t h e system equations. I t . c a n be shown t h a t an nth o r d e r
process cha rac t e r i zed by 5 = Ax + bu is completely c o n t r o l l a b l e i f and
only i f t h e v e c t o r s b , ... Ab, ..- ---- , An'lb a r e l i n e a r l y independent, and com-
- -- -
A ~ , - ' ~ ~ are T T T p l e t e l y observable I f and only i f t h e v e c t o r s c - , A 5 , ---- ' - - l i n e a r l y independent (Kalman, 196%).
The S t a b i l i t y C r i t e r i o n of Popov
The V. M. Popov theorem g ives s u f f i c i e n t cond i t ions f o r t he system
of Equations 2-1 t o 2-3 t o be a b s o l u t e l y s t a b l e . This theorem i s given below:
D I I I D I I I 1 I R I I I i D 1 I 1
13
For t h e system of Emot ions 2-1 t o 2-3 t o be a b s o l u t e l v s tab le + n the s e c t o r TO, K l for the p r i n c i p a l case, and i n the s e c t o r ( 0 , K] f o r t h e p a r t i c u l a r cases i t i s s u f f l c i e n t t h a t t h e r e e x i s t a f i n i t e real number q such t h a t f o r a l l w > 0 -
1 Re[(l + j q w ) C ( j w ) ] + > 0 (2-7)
and, a d d i t i o n a l l y f o r the p a r t i c u l a r ca ses , t h a t t he cond i t ions f o r s t a b i l i t y in- the- l imit be s a t i s f i e d .
A pure ly geometric formulation of t h e Popov theorem can be obtained
from the above a n a l v t i c formulation (Aiaerman and Gantmacher, 1P64). A
modified frequency response, W(jw), is used where Rerw(lu) 1 = RetG(jw) 1 =
X and I m [ W ( j w ) ] = w I m f G ( j u ) ] = Y. Then
Condition 2-7 can now be w r i t t e n as
(2-8) 1 K x-qY + - > 0.
On t h e W-plane t h e l i m i t i n g condi t ion of Equation 2-8 i s the equat ion of a
on t h e real s t r a i g h t l i n e with s lope -which passes through t h e p o i n t - i;: 1 q
a x i s . This l i n e i s c a l l e d t h e Popov l i n e . Condition (2-8) r e q u i r e s t h a t
t he p l o t of t h e modified frequency response l i e e n t i r e l y i n t h e h a l f plane
t o t h e r i g h t of t h e Popov l i n e . Thus t h e geometric formulation of t h e
V. M. Popov theorem is as follows:
I n o rde r t h a t t he system defined by Equations 2-1 t o 2-3 be a b s o l u t e l y s t a b l e i n t h e s e c t o r [O, K] f o r t h e p r i n c i p a l case, o r i n t h e s e c t o r (0, K ] f o r t h e p a r t i c u l a r cases, i t i s s u f f i - c i e n t t h a t t h e r e e x i s t i n the W-plane a s t r a i g h t l i n e , passing through t h e p o i n t on t h e r e a l a x i s with a b s c i s s a - such t h a t t h e modified frequency response W ( j w ) l i e s s t r i c t l v t o t h e r i e h t of i t , and a d d i t i o n a l l y , t h a t f o r t he p a r t i c u l a r ca ses t h e condi- t i o n s f o r s t a b i l i t y - i n - t h e - l i m i t be s a t i g f i e d .
E’
I 1 I 1 I I I I I I I I 1 I i I
11 I I
14
Figure 2-2 i l l u s t r a t e s t h e geometric i n t e r p r e t a t i o n of t he Popov
theorem. I n t h e example of Figure 2-2c, t h e Popov l i n e can be drawn
through t h e o r i g i n . Thus t h e s t a b i l i t y s e c t o r TO, K1 i nc ludes the com-
p l e t e f i r s t and t h i r d quadrants. Figure 2-2d i l l u s t r a t e s a case where
t h e s t a b i l i t y s e c t o r f o r nonl inear systems as found from t h e Popov the-
orem is less than t h e s t a b i l i t y s e c t o r f o r l i n e a r systems.
I n a p r a c t i c a l system, i t i s reasonable t o expect t h e s lope of
the n o n l i n e a r i t y t o b e l imi t ed .
whether t h e s t a b i l i t y s e c t o r of t h e non l inea r system can be increased by
p l ac ing r e s t r i c t i o n s on t h e slope of t h e n o n l i n e a r i t y i n systems such as
t h a t i l l u s t r a t e d by Figure 2-2d. This problem has been i n v e s t i g a t e d and
c r i te r ia developed f o r extending t h e s t a b i l i t y s e c t o r i n such cases bv
Yakubovich (1965a, 1965b), Brockett and W i l l e m s (1965a, 1965b), Dewey
and J u r y (19661, and Dewev (1966). The r e s u l t s of Dewey and J u r v and
Yakubovich are e s a e n t i a l l v t h e same although thev were obtained bv d i f -
f e r e n t approaches. Brockett and Wi l l ems ' r e su l t s are i n a d i f f e r e n t form
and, though they appear t o be more gene ra l , are not as eas i lv i n t e r p r e t e d .
The ques t ion t h e r e f o r e ar ises as t o
Dewey's c r i t e r h n is given he re i n o r d e r t o i n d i c a t e t h e n a t u r e
of t h e r e s u l t s obtained when r e s t r i c t i o n s are placed on t h e s lope of t h e
n o n l i n e a r i t y . The r e s u l t s reported i n t h e o t h e r r e fe rences have t h e same
gene ra l form. Conditions 2-4 (or 2-5) and 2-6 s t i l l apply along with t h e
a d d i t i o n a l r e s t r i c t i o n s t h a t
1. l f < u ) l f M (2-9)
df (6) < K2 2. -K1 < (2-10)
I I . I I I I 1 1 I I I I I I I I 8 I I
15
Figure 2-2a . q > 0 . Figure 2-2b . q < 0 .
Figure 2-2c . Popov Line Goes Figure 2-2d . The Popov S t a b i l i t y
Through the Or ig in . K = 03 . Sector i s Less Than That f o r a
Linear Gain.
Figure 2 -2 .
C r i t e r i o n . I l l u s t r a t i o n s of t h e Geometric I n t e r p r e t a t i o n of t he Yopov
1 I . 1 I
I I I I I I 1 I :I 1 I I 1 P
a
16
The restriction that F(o) be bounded is not required in the other
napers. This restriction makes possible a simpler criterion. The theorem
is as follows:
Theorem: For the system shown in Figure 1, if there exists a finite number q such that for all WLO,
a) 2 2
H(w) = Re[jwqC(jw)] + w c1 + (K2-K1)Re~(jw)-K1K21C(jw)l ,LO 1
b) G(jw)J - K, G ( O ) > - E,
then in the principal case, for all nonlinearitiee with slope restriction (-K states, the response is bounded on io,-) and tends to zero as t + 0 . In the simplest particular case, the theorem remains true for all nonlinearities f(a) in the sector IC, K] such that f(a) -EU is bounded on (--, -) where E > O is arbitrarilp small.
K2) in the sector [O, K) and for all initial 1'
Corollary 1: With the slope restriction f'> -1, condition a) becomes
Re[jwqG(jw)] + w (RcC(jw)-KIIG(jw) I 1 L 0 .
Corollary 2: becomes
Re[ jwqC,(jw) I-w CRec(jw)+ K2)C(jL) I
2 2
With the slope restriction f'> K2, condition a)
2 2 0.
Corollary 3: becomes
With the slope restriction (0, K2), condition a)
2 1 Re[jwqG(jw)] + o (ReG(jw) + -12 0.
Corollary 4: With the slope restriction f'> 0, condition a) be comes
K2
(2-11)
(2-12)
(2-1 3)
(2-14)
(2-15)
(2-16)
Remark: In the particular cases, inequalities 2-11, 2-13, and 2-14 can only be satisfied for the simplest particular case. Inequalities 2-15 and 2-16 can be considered f o r all the parti- cular cases.
The Popov theorp has been extended to time-varving systems bv
Rozenvasser (1963) for the principal case. In the time-varvine systems,
17
f(a) becomes f(a,t). The result is extended to the simplest particular
case by Higgins (1966). The stability criterion is
(2-9 1 1 ReG(jo) + E > 0.
Of the stability theory presented hete, this dissertation makes
use primarily of the V. M. Popov theorem and its extension to time-
varying systems. The extensions of the Popov theory in which the slope
of the nonlinearity is considered are used to indicate how the desim
procedure might be modified in cases where more restrictions on the non-
linearity can be tolerated. It i s noted that in the Popov criterion and
all its extensions, the object in the analysis is to determine the value
of one or more constants vhich indicate the maximum stability sector for
a given G ( s ) .
must s a t i s f y if the system is to be absolutely stable. In this disser-
tation, the approach is to modify G(s) so that no constraints on the
nonlinearity are required in order to s a t i s f y the Popov criterion for
absolute stability.
This determines the constraints which the nonlinearity
CHAPTER I11
CLOSED LOOP DESIGN OF LINEAR SYSTEMS V I A STATE VARIABLE FEEDBACK
In t roduc t ion
The purpose of t h i s chapter i s t o p re sen t a method f o r t h e design
of scalar i n p u t , scalar ou tpu t , l i n e a r c o n t r o l systems v i a t h e s t a t e
v a r i a b l e feedback (SVF) method. F i r s t , t h e procedure f o r designing f o r
a d e s i r e d closed loop t r a n s f e r func t ion i s presented. This procedure is
then used as t h e b a s i s f o r developing a method f o r designinR l i n e a r gain
i n s e n s i t i v e systems which is extended t o c e r t a i n non l inea r and/or t i m e -
varying systems i n Chapter 4.
The procedure f o r designing f o r a des i r ed closed loop response
is formulated from t h e matr ix approach. Th i s procedure i s discussed i n
d e t a i l by Schul tz (1966) from the s tandpoint of the block diaaram. A i -
though t h e b l o c k diagram approach i s more f a m i l i a r t o many c o n t r o l enei-
n e e r s , t h e matr ix approach i s more gene ra l and does not r e q u i r e the manip-
u l a t i o n of t he block diaqram i n t o any s p e c i a l form. When t h e svstem i s
r ep resen ted by a block diagram of t he form assumed by Schul tz , t h e two
methods are equ iva len t .
A f t e r most of t h i s chapter w a s w r i t t e n , i t w a s discovered by t h e
au tho r t h a t Morgan (1963, 1966) has a l s o proposed t h e use of s ta te vari-
a b l e feedback f o r designing l inear svstems t o have a des i r ed closed loop
t r a n s f e r func t ion . H e presents t h e ma t r ix formulation f o r t h e design
18
19
procedure, bu t t h e approach i s somewhat d i f f e r e n t from t h a t presented
here .
v a r i a b l e feedback as c a l l e d f o r by modern c o n t r o l theory and series com-
pensat ion as p rac t i ced i n c l a a s i c a l c o n t r o l theory, therebv combining
t h e advantages of t h e two and inc reas ing t h e v e r s a t i l i t v of t h e desiRn
procedure. While s ta te v a r i a b l e feedback alone can chanqe n e i t h e r t h e
o r d e r of t h e system nor the l o c a t i o n of t h e ze ros , t h e method developed
h e r e can do both.
The procedure presented he re provides f o r a combination of s t a t e
The design procedure makes use of t h e f a c t t h a t t h e closed loop
p o l e s of a l i n e a r system may be forced t o occur anywhere i n t h e s-plane
by feeding back a l l t h e s ta te v a r i a b l e s i n the proper l i n e a r combination
(Brockett , 1965). The requirement t h a t a l l t h e s t a t e v a r i a b l e s b e f ed
back i n d i c a t e s t h a t t he s t a t e v a r i a b l e s should be chosen t o ag ree with
a c t u a l phys i ca l v a r i a b l e s . When i t is no t p o s s i b l e , o r p r a c t i c a l , t o
feed back a l l t he s ta te v a r i a b l e s , t h e ca l cu la t ed va lues of t he feedback
c o e f f i c i e n t s can be used t o determine s u i t a b l e minor loop compensation.
When only t h e output can be fed back, t h e r e s u l t i n g c o n f i w r a t i o n is
similar t o t h a t obtained by the Cuillemin procedure (Truxal, 1955).
The SVF method i s a systematic , completely a n a l y t i c d e s i m
procedure i n which t h e a n a l y t i c a l work is r e l a t i v e l y simple, r e q u i r i n a
t h e s o l u t i o n of n l i n e a r a lgeb ra i c equat ions, where n i s t h e o rde r of
t h e system. It d i f f e r s i n bas i c philosophy from both t h e c lass ica l and
modem design procedures which are i n common use.
i n t h e c lass ica l approach t o t h e design of l i n e a r c o n t r o l systems is t o
modify t h e open loop system i n such a way t h a t when t h e loop is closed
The common practice
20
the performance is satisfactory. That is, the system is comDensated
with a view towards realizing certain open loop characteristics which,
in general, lead to desirable closed loop performance. Among the ex-
tensive literature on this subject are the books by Rower and Schultheiss
(1958) and D'AZZO and Houpis (1960). In the SVF method, the svstem is
compensated so as t o realize a desired closed loop transfer function
which is determined from the performance specifications. Since, ulti-
mately, desirable closed loop response is the goal of the designer, a
method of designing for desired closed loop characteristics provides an
advantage over one of designing for desired open loop characteristics.
Although the motivation for the SVF method of design stems from
a result of modern control theory, the approach is quite different. In
the matrix formulation of modern control theory, a system is represented
by a set of equations as follows:
+ = Ax L L + pl T y " f ' c
(3-1)
< 3-2 ) Here x is an n-dimensional state vector, $ is an (n x n) plant matrix, -. y is the scalar output, c L is an n-dimensional output vector, b is an ,. n-dimensional control vector, and u is a scalar control. In the optimum
control problem the design is based on minimizing a quadratic perform-
ance index of the form OD ca
(3-3) 2 V(x) = (x Ox + pu )dt = I,
The well h o r n solution to this problem (Kalman, 1 9 6 4 ) is that the optimal
control is T u=-&c. ( 3 - 4 )
2 1
That i s , t h e optimal c o n t r o l c o n s i s t s of a l i n e a r combination of a l l the
s t a t e v a r i a b l e s .
reduced Y a t r i x R i c a t t i equation,
This l i n e a r combination may be found by solvine. t h e
T -1 T T A R + R A - R b p b R + r r " 0 ; - -0 -0 - -0- - -0 - - k i s then found from k = Rob.
t i o n q u i t e d i f f e r e n t from t h a t of t h e c lass ical method of s e r f e s compen-
s a t i o n with u n i t y feedback from t h e output. I n f a c t , u s i n e t h e frequency
domain c r i t e r i o n f o r op t ima l i ty as developed by Kalman (1964), i t can be
shown t h a t very few systems designed bv t h e classical method are optimal
f o r any q u a d r a t i c performance index. Systems desietned by t h e SVF method
may or may no t s a t i s f y thris c r i t e r i o n . This is discussed f u r t h e r i n t h e
s e c t i o n on ga in i n s e m i t i v e systems.
This r e s u l t suggests a system configura- - - - -
A major d i f f i c u l t y i n the optimum c o n t r o l approach t o desiRn
arises from t h e l a c k of s u i t a b l e c r i t e r i a f o r spec i fy ine t h e performance
index.
by Equation 3 - 4 , bu t t h e design c r i t e r i a i s a des i r ed closed loop t r ans -
f e r func t ion r a t h e r than a performance index.
The SVF method makes use of t h e system conf igu ra t ion sueeested
The system configurat ion r e s u l t i n g from SVF design i s i l l u s t r a t e d
i n Figure 3-la. G (9) r ep resen t s t h e p l a n t t o be controll-ed, C, !s) t h e
series compensation, and K an unspecif ied l i n e a r Rain. G(s) i s def ined
as G(s) = Gc(s)G ( 8 ) and KG(s) is t h e forward t r a n s f e r func t ion . Since
i t is u s u a l l y d e s i r a b l e t h a t kl = 1 i n o r d e r t h a t t h e output w i l l follow
t h e i n p u t with as s m a l l a s teadv s t a t e e r r o r as p o s s i b l e , t h i s value is
used throughout t h i s d i s s e r t a t i o n . The loss of t h i s v a r i a b l e parameter
i s compensated f o r by providinq t h e unspecif ied gain, K , precedinp: C ( s ) .
P C
P
2 3
Since the SVF method requires t h a t t h e s t a t e v a r i a b l e s correspond
t o a c t u a l phys i ca l v a r i a b l e s which can be measured and fed back, i t some-
times happens t h a t a term involving 6 must be included i n Equation 3-1.
For example, i n Figure 3-lb, gn = A x + aKu + KG, and Equation 3-1 becomes n
(3-5a)
o r
% = Ax - - + Bu, - - (3-Sb)
where
w i t h b - and d - n-dimensional vec to r s and
U 7
Systems represented by Equation 3-1 w i l l be r e f e r r e d t o i n t h i s d i s s e r -
t a t i o n as Class I and those represented by Equation 3-5a or 3-5b as w
Ciass 11.
The remainder of t h i s chap te r i s arranged i n the following
o rde r :
1. The ma t r ix formulation of SVF design i s developed.
2 . A design procedure f o r gain i n s e n s i t i v e systems i s developed.
3 . The procedure t o be followed when a l l t h e s t a t e v a r i a b l e s
cannot be f ed back i s discussed.
4 . The r e s u l t s a r e summarized.
2 4
The SVF Method - ,Ma t r ix Formulation
The genera l procedure is developed on the b a s i s of Class I1
systems s i n c e the r e s u l t s can e a s i l y be extended t o Class I systems bv
l e t t i n g + = 0.
From Figure 3-lb,
u = -& T 5 + r. (3-6)
This i s t h e same as Equation 3-4 except f o r t he term r , which r ep resen t s
t h e s c a l a r input t o t h e c losed loop system. S u b s t i t u t i n g Equation 3-6
i n t o Equation 3-5a, transforming, and so lv ing f o r X ( s ) gives - X(s) = [a! + dk T s-(A-bkT)]-'(5 + d S ) R ( s ) . - - - -.* (3-7)
This is combined with Equation 3-2 t o g ive the closed loop t r a n s f e r
func t ion i n terms of f a c t o r s of t h e form Kki (From FiRure 3-1, i t i s
ev ident t h a t K is a f a c t o r i n each element of and cj ) .
(3-8)
This t r a n s f e r func t ion can now be compared wi th t h e d e s i r e d t r a n s f e r
func t ion , and t h e n a l g e b r a i c equat ions t h a t r e s u l t from equa t ine cor-
responding c o e f f i c i e n t s of s can be solved f o r t h e Kk With t h e pre-
v ious assumption t h a t kl - 1, K and t h e o t h e r ki can then be found.
Actua l ly , on ly the denominators of t h e two t r a n s f e r func t ions need t o
be compared i n order t o determine the k and K , a s state v a r i a b l e feed-
back does not a f f e c t t he zeros of t h e t r a n s f e r func t ion! i . e . , t h e zeros
of t h e closed loop t r a n s f e r func t ion a r e t h e same as t h e zeros of t h e
open loop t r a n s f e r func t ion . This can be shown from t h e equiva len t
system of Figure 3-lc. The expression f o r t h e eouiva len t feedback
1'
i
25
transfer function, H ( s ) , is derived in the section on gain insensitive
systems, and it is shown there that the Doles of H (6) are al.so zeros eq
eq of C ( s ) . Thus the zeros of the closed
KC(s) = l+KG(s)H (s) ’
eq
must be the same as the zeros of G ( s ) .
loop transfer functjon,
.. If the numerator of C(s) is not
Y equal to the numerator of the desired $s), series compensation is neces-
sary to make the two compatible.
not necessary is called the simplest case, and the case where series
The case where series compensation is
compensation is required is called the eeneral caee.
If G(s) is not known, the numerator must be found in order to
compare it with the numerator of the desired closed loop transfer func-
=. Transforming Equations 3-2 and tion. From Figure 3-1, KC(s) - 3-5, combining, and solving for
U(S)
u(s) gives
This can be written in the form
where p(s) is the matrix [SI-A] . . . I and pa(s) is the adjoint of ?(SI. the adjoint has no poles (only positive powers of 8 occur in the matrix),
it follows that the poles of the transfer function must be zeros of det
Since
?(SI. of the determinant may be cancelled by zeros in the numerator.
The converse does not necessarily hold, since one or more zeros
The
necessary and sufficient conditione that the converse hold are that the
system be both controllable and observable (Brockett, 1 9 6 5 ) . It is not
26
a usually necessary t o f i n d a l l t h e elements of P (3) i n o rde r t o determine
t h e numerator of G ( s ) .
- For example, i f t he non-zero elements of $ are
designated c and the non-zero elements of (z are d e s i m a t e d d the onlv
elements of t h e a d j o i n t matrix t h a t a f f e c t t h e numerator of G ( s ) are t h e i j'
elements i n t h e ith rows and the jth columns. I n systems where y = xl,
t he only non-zero element of c i s c Therefore!no elements i n any row 1'
o t h e r than t h e f i r s t a f f e c t t h e numerator of G(s). I f i t i s d e s i r a b l e t o
determine the complete inverse matr ix , t h e Leve r r i e r a lgori thm
(Gantmacher, 1959) provides an o r d e r l y procedure f o r t h e simultaneous
computation of t h e c o e f f i c i e n t s of t h e c h a r a c t e r i s t i c polynomial and t h e
a d j o i n t matr ix , and i s adaptable t o machine computation.
Equation 3-8 i s now w r i t t e n i n the form
(3-11)
T T a where i s the ma t r ix [s (z + @ ) - (4 - bk ) I and F (s ) is the a d j o i n t
of Ek(s).
i n a t o r of Equation 3-11 m u s t b e determined i n o r d e r t o f i n d -(SI i n terms
of the ki and K, and t h i s is equal t o d e t Fk(s) .
-K
Assuming t h a t t h e numerator of G(s) i s known, only t h e denom-
Y R
The above d i scuss ion i n d i c a t e s t h e general procedure of SVF
desiEn. Before o u t l i n i n g t h e s p e c i f i c procedure i t i s shown t h a t t h e
a l g e b r a i c equat ions which must b e solved f o r t h e Kk are always l i nea r .
The proof makes use of t he following theorem (Nering, 1963):
i
I f A ' i s t h e ma t r ix obtained from A by adding a m u l t i p l e of one row ( o r column) t o another, then d e t A' = de t A .
2 7
Since K i s a f a c t o r of t h e elements of both b and d , t hese v e c t o r s are
w r i t t e n as
- -
- bl
b - :
b -
n Kb' n
The ma t r ix of Equation 3-8 has t h e form
c 1 7 -l
is(l+Kk d')-all+Kk b ' l lsKk d'-a +Kk b ' I 1 1 1 1 1 2 1 1 2 2 1 1
IKk d's-a21+Kklb; 1 s(l+Kk d')-a22+Kk b ' 1 2 I 2 2 2 21
L A I - -I
jKk d' s-a +KklbA I 1 Kk2dAs-an2+Kk2bA 1 - 1 n nl L -.i L -4
-
...
...
r
Kkn d -a 2n+Kknb 1
i s ( l+KkndA) -ann+Kknbnj J L
-
where t h e a are t h e elements of t h e A matrix.
s d i + b;
Now t h e nth row t l
i s sub t r ac t ed from the f i r s t row. I f 6 = 0 i j by sd' + bA
n
f o r i p j and 1 f o r i = j , then t h e j t h element of t h e f i r s t row i s
given by t h e fol lowing expression:
t h This process is repeated f o r each of t h e f i r s t (n-1) rows, w i th t h e n
28
sd; + b; t h row mul t ip l i ed by sd,: + b:, being s u b t r a c t e d from the i row. Then the
t h j th element of t h e i row is Riven by t h e following expression:
sd; + b; sd; + b; L i j - " n j sd; + b i - a i j + anj sd; + b; (3-12) 1
According t o t h e above theorem, t h e determinant is no t changed by t h e s e
ope ra t ions . Since Kk is no t a f a c t o r i n the terms of Equation 3-12, i t
follows t h a t t h e f a c t o r s Kk
i
appear only i n t h e nth row of t h e new matr ix . i
Thus t h e determinant contains only f i r s t o rde r terms i n t h e Kk and has
the form i
(Kkl , Kk,) sn-' -+ ' +f (Kkl , , Kk,) s + fn - l (3-1 3)
where the func t ions fo,-**,f are l inear i n t h e Kk n t' of t h e des i r ed closed loop t r a n s f e r func t ion can be w r i t t e n i n t h e form,
The denominator
sn-l +' -+ P s + Po. n P ( s> = s + Pn-l 1 (3-14)
Equating corresponding c o e f f i c i e n t s of Equations 3-13 and 3-14 Rives a
set o f n l i n e a r a l g e b r a i c equations which are l i n e a r i n Kki and can b e
solved f o r K and k (k
so only n unknowns occur i n the n equat ions) .
i s usua l ly set equal t o uni ty as noted p rev ious ly , - 1
I n t h e general case, i t is n o t r equ i r ed t h a t t he numerators of
G (s ) and ~ ( s ) be compatible, al though -(s) R may never have a Dole-zero
excess less than t h a t of C (s), as t h i s would r e q u i r e a compensator i n
which the numerator is of higher o rde r than t h e denominator.
Y Y P
P Series
I I I I I I I I I u 1 I I I I I I I I
L
29
compensation is used i n add i t ion t o s ta te v a r i a b l e feedback i n o r d e r t o
realize t h e d e s i r e d response. This u sua l ly means adding pole-zero p a i r s ,
w i th t h e zeros being r equ i r ed t o shape t h e closed loop frequency response
curve o r t o provide t h e required v e l o c i t y e r r o r c o e f f i c i e n t (Truxal,
1955).
The zeros of t h e compensator are assumed t o be known s i n c e ,
Y u n l e s s they cancel a po le i n G (s), they w i l l a l s o be zeros of ,(SI.
Each pole-zero p a i r i n c r e a s e s the o rde r of t h e system, except i n those
cases where poles o r zeros are cancel led, and t h e number of s t a t e
v a r i a b l e s by one. This means t h a t each pole-zero p a i r adds two new
parameters , the p o l e l o c a t i o n and t h e feedback c o e f f i c i e n t . One of
t h e s e must be chosen a r b i t r a r i l y and t h e o t h e r determined along with
t h e o t h e r ki.
series compensator r e s u l t s , and t h e l o c a t i o n s of t h e poles must be
determined along with t h e value of t h e non-zero k
l o c a t i o n s are chosen, t he a d d i t i o n a l feedback c o e f f i c i e n t s must be
determined along wi th t h e feedback c o e f f i c i e n t s froin the or ig i sa l q s t e m .
On t h e b a s i s of t h e above d i scuss ion , t h e following design
P
I f t h e new feedback c o e f f i c i e n t s are chosen t o be zero, a
I f t h e po le i'
procedures are suggested:
The Simplest Case
1. Describe t h e system i n terms of meaningful, phys i ca l s ta te
v a r i a b l e s and assume t h e s e are a l l a v a i l a b l e and are f e d
back through constant gain elements.
Choose t h e des i r ed closed loop response, ,(SI. Y 2.
3. From EQUatiOn 3-8, f i n d Y (e) i n terms of K and t h e ki.
I I I I I I I I I I I I I I I I I I I
30 Y R
4 . Equate t h e expressions f o r t he denominator of -(s) from
s t e p s 2 and 3 and solve f o r K and the k by equat ing l i k e
powers of s .
i
5. Use t h e known values of t h e system parameters t o r e a l i z e a
f i n a l system configurat ion. I f a l l t h e s t a t e v a r i a b l e s are
no t a v a i l a b l e t o feed back, use the ca l cu la t ed va lues of k
t o determine s u i t a b l e minor loop compensation.
The General Case
1. Same as s t e p 1 of the s imples t case.
2. Same as s t e p 2 of the s imples t case.
3. Add a s u f f i c i e n t number of pole-zero p a i r s t o make G ( s ) =
Y G c ( s ) G (s) compatible with the des i r ed ~(3).
number added i s p.
Assume t h e P
4 . Choose p a r b i t r a r y pole and/or feedback c o e f f i c i e n t s
a s soc ia t ed wi th the p new s ta te v a r i a b l e s introduced.
5 . Same as s t e p 3 f o r t h e s imples t case.
6 . Eqiiate t h e expressions f o r the denominator of -(SI from s t e p s
2 and 5 and so lve fo r K and t h e k by equat ing l i k e powers
of 9.
Y, R
i
7. Same as s t e p 5 f o r t he s imples t case.
The fol lowing examples i l l u s t r a t e t h e design procedures. Example
3-1 rep resen t s t he s imples t case of a Class I system. Example 3-2
r ep resen t s t he genera l case of a Class I1 system.
Example 3-1: I n Figure 3-la, l e t C: (s) = 1, and assume t h a t t h e system
i s represented by t h e set of d i f f e r e n t i a l equat ions
C
I 1 . I I 1 I I i I I I I I I 1 1 I 1 I
31
k1 = x2
2, =-x2 + 2x3
ir = -4x + 5Ku
T u = - k x + r T y = c x
3 3
5 -
= x1 - 5
kl = 1.
Comparing t h i s with Equation 3-1 y i e lds
0 1 0
A = - 0 - 4
l l : = 0 O I
J.
Y The des ired $s) is chosen as
Y 260 = 260 (3-15) $ 5 ) =
(s2 + 4s + 13)(s + 20) s3 + 24s2 + 93s + 260
From the A matrix and the b vector, the matrix of Equation 3-8 is found - ..,
to be
-1 0
s + l -2 r s l o I 1 . 1 5K 5Kk2 s + 4 + 5 Kk3
I I- -
Substituting t h i s matrix and the ! and s vectors into Equation 3-11 with
$ = gives
10K - 3 s +(5+5Kk3)s 2 +(4+5Kk3+10Kk2)s+10K
(3-16)
I I * .
I I I I 1 1 I I I 1 I I I 1 1 I 1
32
Equating Equation 3-16 t o t h e des i r ed t r a n s f e r funct ion of Equation 3-15
gives t h e following set of l i n e a r a l g e b r a i c equat ions:
10K = 260
5+5Kk3 = 24
4+5Kk3+10Kk2 = 93.
These equat ions can be solveir f o r Kki, giving
K = 26
19 Kk3 = 5
Kk2 = 7.
These equat ions are e a s i l y solved f o r t h e proper set of feedback coe f f i -
c i e n t s t o give t h e d e s i r e d closed loop t r a n s f e r funct ion. The r e s u l t is
given below:
K = 26
k2 =I 0.269
k, = 0.146. d
Example 3-2: Assume t h a t t h e system t o be c o n t r o l l e d is represented by
t h e equations.
icl = x2
ic 2 = -2x2+u
T u = -k xsr T - 4
Y " $ X ' X 1
kl = 1.
For t h i s system,
33
(3-17) 0 :+*I - - A
It i s assumed t h a t t h e des i r ed closed loop t r a n s f e r funct ion is
Y 5 (s+4) R(S) = 2
(6 +2s+2) (s+10)
The r e a l i z a t i o n of t h i s des i r ed t r a n s f e r func t ion r e q u i r e s , i n a d d i t i o n
t o s ta te v a r i a b l e feedback, a compensator of t h e form as shown i n
Figure 3-2a with t h e value of e i t h e r a o r t h e feedback c o e f f i c i e n t k3 t o
b e chosen a r b i t r a r i l y . I f a i s chosen t o be 10, t h e system equat ions
become
S+a
5 = x 1 2
k2 = -2x +x 2 3
k3 = -l0x3+4ku+k;
T u = -k fir “ 5
T
1 y = = x
kl = 1.
Comparing these equat ions with Equation 3-5 le3ds t o
b =
1
..
0 7
“ I d = 0 c = “ 1 0
o i
From t h e A matrix and t h e b and d v e c t o r s , t h e ma t r ix of Equation 3-8
is found t o be
“ 5 5
I - . 34
Figure 3-2a. The Plant of Example 3-2 with S e r i e s Compensation.
i
1 y = x Gp(S) . K s + 4 x3
s + 10 w
-[ k3 I -p2 1
Figure 3-2b. Method of Controlling the Plant of Example 3-2.
I ' . - S
' I
-1
-
0
0 s+2 -1
3 Ks+4K Kk s +4Kk s+Kk 3s+10+4Kk -
35
Substituting t h i s matrix and vectors b , c , and d into Equation 3-11 gives - - . -.
(3-19) Y K (s+4) ,(SI = (l+Kk3)s 3 +(12+6Kk3+Kk2)s 2 +(20+8Kk3+4Kk2+K)s+4K
Equating Equation 3-19 t o the desired transfer function of Equation 3-18
gives
4 K - = 20 1+Kk3
1 2+6Kk3+Kk2 = 12. 1+Kk3
The so lut ion t o t h i s set of equations i s
K = 4.
This completes the design procedure. The f i n a l system configuration is
shown i n Figure 3-2b.
36
Applicat ions i n Nuclear Reactor Control Sys t em Design
I n t h i s s e c t i o n several examples of t h e a p p l i c a t i o n of s t a t e v a r i a b l e
feedback design t o r e a c t o r con t ro l a r e presented.
method makes use of a l l the s y s t e m v a r i a b l e s , i t i s p a r t i c u l a r l y amenable
t o the c o n t r o l of mult i region r e a c t o r s as w e l l as s i n g l e region models.
With the recent emphasis on s p a c i a l k i n e t i c s , t h i s method of c o n t r o l i s
apropos.
Two Temperature Region Reactor
Since t h i s new design
The block diagram for the l i n e a r two-temperature region r e a c t o r with
s t a t e v a r i a b l e feedback c o n t r o l and neg lec t ing delayed neutrons, i s shown
i n Fig. 3.3 where
a = i
ki =
K =
Ki =
a = i
x1 =
x2 =:
x = 3
- x4 -
B12 -
B 2 1
-
a
Y '
h e a t removal c o e f f i c i e n t s of ith region
feedback c o e f f i c i e n t s
gain constant of c o n t r o l l e r
p r o p o r t i o n a l i t y cons t an t between power and
t h temperature of t he i region
temperature c o e f f i c i e n t of
ith region
neutron densi ty o r power
temperature i n region 1
temperature i n region 2
r e a c t i v i t y o f t h e
r e a c t i v i t y input from c o n t r o l l e r
temperature coupline, c o e f f i c i e n t from region 1 t o 2
temperature coupling c o e f f i c i e n t f rom region 2 t o 1
r e c i p r o c a l of t h e c o n t r o l l e r tize cons tan t
37
I I
I I i
Figure 3 . 3 Block Diagram of Two Temperature Region Reactor w i t h State Variable Feedback Control
I I I
38 I
1
n = steady s ta te neutron density or power 0
T = ef fect ive neutron generation t i m e
The d i f f erent ia l equations defining the system are
-n n n 0 0 0 a x - - u x + - = - % T 1 2 T 2 3 T x 4
i 2 = K x - a ( x 1 1
k3 = K x -a (B
1 2 + R 2 1 3 x )
x + x3) 2 1 2 1 2 2
ir4 = -YX + KU 4
Referring to the equations above, the terms i n Eq. (3-8) are given by
- 0
0 b =
0
where d = 0 .
Then
T [SI-A-bh ] = - - -..
-n 0 -
T a1
1 -a
-a B 2 12
0
K -
n - ' a ' t . 1
1 s+a
a2B12
%2
- -n n
0
T "2 T - 0 -
-a B 0
0
1 21
2 -a
kT = kl k2 k3 k4
n - ' a ' t . 2
a B 1 21
2 s+a
Kk3
0
s+y+K k 4
= F ..
39
Letting the co-factors of F be F and the determinate of F be det F, the
inverse of F is ij .. -,
I
and Eq. (3-8) becomes
%d. P cTF-\ R(s ) - ., -
After some matrix algebra, the transfer function for the case in point is
m=- KF41 R ( s ) det F
and in terms of the
and
sys tern parameters
n 0 2 - - [s +(a +a )s+ala2(l-B B ) ] F41 T 1 2 12 21
det p = s4(al+a2+y+k4K)s 3 +[ (al+a2) (y+Kkq)+ala2(1-B12B21) + K2a2no
n 2 0
‘I T 12 21 T
K u n n + klK]8 +{ala2(y+k4K)(1-B B ) + - [-aialB21K2 +
n T 1 1 2 1 2 2 1 2 1 2 2 1 1 1 2 + A{ [-a a B K -u B K +K a a +K a a ](rck4K)+Klk3Ka2B12+klKa1a2
+a B K k K+klK(a1a2)B12B21+K1k2Ka2+K2k3Kal) 1 2 1 2 2
As in the previous example it is seen that the zeroes of the system transfer
function are independent of k
ki.
by selecting proper values for the feedback coefficients k
while the pole locations are a function of i’
Therefore the form of the system time response can be chosen at will
i’
40
Example 3-3: As an example, consider a system t h a t has the following
cons t an t s . -5 K = 2x10 degrees/watt sec 5
1 n = 10 watts 0
-1 y = 10 sec K2 = degrees/watt sec
-1 a = 0.01 s e c K = 1.0
a2 = 0.05 sec
1
-1 T = 0.1 sec
a = per degree B12 =: -0.2
u2 = 10 pe r degree B21 = -1.0
1
-4
For these va lues of t he system parameter
= 10 6 2 (s + 0 . 0 6 ~ + 4 ~ 1 0 - ~ ) F4 1
Therefore t h e system has two zeroes c l o s e t o the o r i g i n . Suppose t h a t t he
des i r ed dynamics of t he system i s Riven by t h e second o r d e r t r a n s f e r func t ion
l o 6 [q d = s2+20s+200
which has well-behaved t r a n s i e n t c h a r a c t e r i s t i c s with a dampeninR r a t i o of
0.707 and d e s i r a b l e frequency response.
c h a r a c t e r i s t i c s , Eq. 3-8 must equal
To rea l ize t h e s e d e s i r e d system
10 6 2 (s + 0 . 0 6 ~ + 4 ~ 1 0 - ~ )
R(s) (s2+20s+200) (s 2 +o.O6~+4xlO-~)
o r
+0.06~+4~10-~) 2 +201s +12s+0.08
Equating t h e c o e f f i c i e n t s of l i k e powers of s i n t h e denominator of t h e
equat ion above t o d e t F and so lv ing t h e r e s u l t a n t l i n e a r a lRehra i c simul-
taneous Eqs. f o r ki y i e l d s
5
41
m ' 0
IC) cv 0 0
0 In -
<u x
42
0 o_
0 10
k 0 U
0 cc E 0
f
a
c
-4 kl = +1.98x10
k, = -0.02 ‘-
4 3
Figures 3 . 4 and 3.5 give t h e response of
k3 -0.002
k4 = +10.0
t h e system v a r i a b l e s t o a s teo
demand i n power. From t h e response curve of xl, i t is seen t h a t t h e t r ans -
ient response behavior corresponds e x a c t l y t o t h a t expected from the d e s i r e d
system t r a n s f e r funct ion. Further from s imula t ion s t u d i e s , v a r i a t i o n s i n
t h e feedback c o e f f i c i e n t s k
v a r i a b l e s , had v i r t u a l l y no e f f e c t on t h e system dynamics.
and k correspondinR t o t h e temperature s t a t e 2 3
Neglecting t h e s e
two feedbacks e s s e n t i a l l y d i d n o t a l t e r t h e s t e p response.
i n t he c o n t r o l rod p o s i t i o n feedback c o e f f i c i e n t had no n o t i c e a b l e e f f e c t on
A f. 20% change
t h e t r a n s i e n t response. S e t t i n g t h i s feedback cons t an t equal t o zero gave a
t r a n s i e n t response with a damping r a t i o of about 0.25. Changes i n t h e s t eady
s ta te power level r e s u l t i n g from a s t e p demand i n power were sensitive t o
v a r i a t i o n s i n k
decreased, t h e s teady state power level inc reased and the system became more
the ou tpu t s ta te v a r i a b l e feedback c o e f f i c i e n t . A s kl 1’
damped. For i n c r e a s i n g va lues of kl the s t eady s ta te power level decreased
and t h e system became less damped. From t h e above i t can be concluded t h a t
bbr t h e case i n p o i n t only t h e two feedback c o e f f i c i e n t s kl and k
c a n t i n determining t h e system dynamics.
are s i g n i f i - 4
Coupled Core Reactors
Another good i l l u s t r a t i o n of the a p p l i c a t i o n of t h i s new design technique
is the c o n t r o l of a coupled core r e a c t o r , a block diagram of which, along
wi th the feedback c o e f f i c i e n t s , i s shown i n Fig. 3.6 . For convenience,
delayed neu t rons have been neglected, and t h e cores are assumed t o b e i d e n t i c a l
w i t h t h e same neutron coupling c o e f f i c i e n t . I n this model t h e symbols are as
denoted i n t h e previous example except
r
I I
I 1
I * 3 D s+
I K ! I 1 7
n
Figure 3 . 6 Block Diagram of Coupled Core Reactor with State Variable Feedback Control
i 45
x = neutron dens i ty o r power i n core 1 1
x2 = temperature i n core 1
x3 = neutron dens i ty o r power i n core 2
x4 = temperature i n core 2
D = neutron coupling c o e f f i c i e n t
y = t o t a l neutron dens i ty o r power of combined cores
x5 = r e a c t i v i t y i n p u t from c o n t r o l l e r
The d i f f e r e n t i a l equat ions f o r t h i s system a r e
i
7
I n t h i s example
rn
[SI-A-bkl] - I - - =
i
an n j, = - - D D - - 0 x + - x 0 1 T X 1 + T X 3 T 2 T 5
G2 = Klxl-ax2
an D 0 x 3 + - x - - 2 3 = - - T T I T x4 D
k4 = K1x3-ax4
?5 = -YX +Ku 5
g = x + x 1 3
an 0 -
T s s
T
s+a 1 -K
-Dn 0 0 -
T
0 0
Kkl Kk2
-D - 7
0
D S4-
T
-K1
Kk3
n 0 - - 0 T
0 0
an 0 0 -
T
s+a 0
Kk4 s+y+Kk
= F -.
46
and the c losed loop
where aga in F are
of F. i j
-
t r a n s f e r function i s
t h e co-factors of F and d e t F denotes the determinate 5 -
Example 3-4: To demonstrate the above, assume t h e following va lues
f o r t h e system parameters
K1 = degrees/watt sec 5 n = 10 w a t t s 0
a = sec-l T = 0.1 sec
a = 10- p e r degree 3 D = 10- 3
y = IO sec- l K = 1.0
and s p e c i f y t h e same system t r a n s f e r func t ion as i n t h e previous example.
To r e a l i z e t h i s d e s i r e d system t r a n s f e r func t ion Eq. (3-8) must equal
6 2 m, 10 (s+O. 01) (s +10r)Os+10) 2 2 R(s) (s+O.Ol) ( 8 +10OOs+10) (s +20s+200)
-. 1 roilowing the s a m e procedure as before equat ing t h e c o e f f i c i e n t s of iiice
powers of s i n t h e denominator of the equat ion above t o t h e corresponding
c o e f f i c i e n t s of powers of s i n the denominator.of d e t F and s o l v i n g t h e
l i n e a r a l g e b r a i c simultaneous equations thus formed, t h e va lues f o r k are
determined.
- i
k3 = +2 X -4 kl = +2 X 10
k4 = -1.92 X -2 k2 = -2 X 10
k5 = +10
Figure 3 . 7 shows t h e response of the system t o a s t e p demand i n power f o r
a coupl ing c o e f f i c i e n t D = 0.001, 0.01, and 0.1. As i n d i c a t e d i n t h e f i g u r e ,
- ' X
- 0
0
0 (I
0 cj
h
W
w x H w
. . aJ 0 V
'cf F l3 E a, a
48
t h e des i r ed system response y ( t ) is i n s e n s i t i v e t o changes i n D.
t h e response of t he i n d i v i d u a l cores w i l l depend on D.
Obviously,
As i n t h e previous example, the temperature feedbacks had l i t t l e
in f luence on the system dynamics and neg lec t ing them made only a very
s l i g h t change i n the system response. Again t h e f i n a l va lue of t h e system
power response
i n c r e a s i n g f o r
p o s i t i o n s t a t e
damping r a t i o ,
w a s a f f e c t e d by v a r i a t i o n s i n the power feedback c o e f f i c i e n t s ,
decreasinR va lues i n kl o r k3. Neglecting the c o n t r o l rod
v a r i a b l e feedback y i e l d s a s t e p response wi th about a 0.3
and again v a r i a t i o n s i n k5 f. 20% had l i t t l e e f f e c t on t h e
power response. It can be concluded the re fo re t h a t only the feedback co-
e f f i c i e n t s kl, k and k have apprec iab le e f f e c t s on t he system behavior. 3 5
I n a c c e s s i b l e S t a t e Variables
I n t h e examples presented , i t w a s assumed t h a t a l l s ta te v a r i a b l e s were
a v a i l a b l e . I f delayed neutrons a re included i n t h e system model, obviously
i t is impossible t o measure neutron precursor concent ra t ion f o r c o n t r o l pur-
poses , and thus t h i s s t a t e v a r i a b l e i s not a v a i l a b l e . However, i t can e a s i l y
be generated provided i t can 5 e detemdried from a mathematicai r e l a t i o n s h i p .
To demonstrate, consider t h e block d iag ram of a r e a c t o r system with delayed
neutrons and s ta te v a r i a b l e feedback, a s shown i n Fig. 3.8a.
Obviously, the state v a r i a b l e x1 cannot be measured; however, i t can
be generated by moving t h e l i n e a t x t o x as shown i n Fig. 3.8b. Fig. 3.8b 1 2
reduces t o the form i n Fig. 3 . 8 ~ .
Clear ly , from the d iscuss ion above, s ta te v a r i a b l e s t h a t are no t ava i l -
a b l e can be generated by p l ac ing a frequency dependent element i n the feed-
back path as demonstrated. I f , f o r some reason, one of t h e s t a t e v a r i a b l e s
cannot be fed back o r cannot be generated, then only n-1 poles of t h e des i r ed
closed-loop t r a n s f e r func t ion can be s p e c i f i e d , where n is the o rde r of t h e
system, the o t h e r po le f a l l s where i t may.
49
I I
L I
Figure 3.8 Block Diagram of State Variable Generation
50
Gain I n s e n s i t i v e Systems
,
A system designed by t h e method proposed i n t h i s s e c t i o n is
shown t o possess two i n t e r e s t i n g c h a r a c t e r i s t i c s : 1) The closed loop
t r a n s f e r func t ion is e s s e n t i a l l y independent of a ga in K loca ted as
shown i n Figure 3-lb. This is t h e c h a r a c t e r i s t i c t h a t l eads t o t h e
design of c e r t a i n nonl inear and time-varying systems i n l a te r chapters .
2) The frequency domain c r i t e r i o n f o r an optimal con t ro l s u b j e c t t o a
quadra t i c performance index of the form of Equation 3-3 is always satis-
f i e d f o r some performance index, As mentioned previous ly , c o n t r o l systems
i n genera l do not have t h i s property.
The second i t e m is considered f i r s t . A frequency domain crite-
r i o n f o r an optimal c o n t r o l subjec t t o a performance index of t h e form
ofEquation 3-3 has been shown by Kalman (1964) t o be ( f o r a c o n t r o l l a b l e
s y s tem)
ll+kT@(s)k12 - - = 1+ITT$(s)"12 . (3-20)
Here 9 ( s ) = [SI-&]-~ is t h e reso lvent mat r ix of t h e p l a n t t o be con-
t r o l l e d p l u s any series compensation.
P f o r t h e open loop t r a n s f e r funct ion $s) = KG(s).
Equations 3-1 and 3-2 are solved
The r e s u l t i s
From Figure 3-lc
Heq(s) =
and Equation 3-21,
P
(3-21)
51
Equation 3-20 can now be w r i t t e n i n the form
(3-23)
o r
Thus f o r a system t o be optimal f o r some q u a d r a t i c performance index,
i t is necessary t h a t i nequa l i ty 3-24 be s a t i s f i e d . This is r a r e l y the
case i n a system designed i n the c l a s s i c a l manner, as can b e seen from
Figure 3-9, where curve a represents a t y p i c a l open loop func t ion ,
KG(s)H ( 8 ) . I n o r d e r t o be an optimum system, t h e p l o t of KG(s)I! (s)
must remain o u t s i d e t h e u n i t c i r c l e wi th c e n t e r a t -1. Such a system is
represented by curve b.
method of t h i s s e c t i o n always s a t i s f y t h i s condi t ion .
eq eq
It is shown below t h a t systems designed by the
S ince the des ign procedure of t h i s s e c t i o n is extended t o t h e
case where the ga in K is nonl inear and/or time-varying i n Chapter 4 , i t
is d e s i r a b l e t h a t K no t appear i n t h e t r a n s f e r func t ions used to der ive
the express ion f o r H ( 8 ) . In o rde r t o accomplish t h i s , u' is def ined
as shown i n Figure 3-lc (u' = Ku) and t h e r e l a t i o n s h i p b = Kb' is used. eq
Equation 3-21 is rep laced with
and Equation 3-22 wi th
_kT9 ( s )h ' c @ ( s ) b '
H e q ( s ) =I
-., ..
Equation 3-22' is now w r i t t e n i n the form
(3-21)
(3-22')
52
t
- P o l a r P l o t of ari Optimum System
= G ( j w ) f o r a S e r i e s Compensated Unity Feedback System.
Figure 3 - 9 . Nyquist Diagram of GH (jo) f o r a System t h a t Would be
Considered Satisfactory i n Terms of Conventional Cri ter ia .
eq
5 3
(3-25) detF (s) kTFa(s)bl - -
cTFa(s)b' B (SI = 5
cTFa(s)b' ..- - - eq
detF(s) .. where Fa(s) is t h e a d j o i n t of t h e r e so lven t matrix F ( s ) . The elements
a o f F (s) conta in terms of order (n-1) and lower i n t h e numerator and
have no poles . I f a l l t h e k a r e non-zero, i t follows t h a t t h e numera-
t o r of H (9) w i l l have (n-1) zeros and, s i n c e t h e denominator of
Equation 3-25 is t h e numerator of G ( s ) , t h a t t h e po le s of H (s) are
equal t o zeros of C(s) .
and Equation 3-25 becomes
..
i
eq
eq For Class 11 systems, X(s) = [sI-A]-l(b'+d's)u'(s) . . I
5 -
kTFa(s) - 5 (b'+d's) - - sTFa(s) (b'+d's) . . . . He,(s> = (3-26)
This is given h e r e i n o r d e r t o show t h a t f o r a system conf igu ra t ion such
as t h a t of Figure 3-UJ,where b ' and d ' are r e l a t e d by a cons t an t , t h e r e
w i l l be a c a n c e l l a t i o n i n Eq. 3-26 and H
5 - ( 5 ) w i l l not have a p o l e
eq 1 corresponding t o t h e ze ro of G(s) a t s = - - ,
The des i r ed H (s) is chosen so t h h t t h e (n-l) zeros are eaua l t o
(n-1) of t h e n po le s of G ( s ) . From t h e above d i scuss ion , i t then follows
1 1
t h a t
(3-27)
except f o r a system configurat ion such as t h a t shown i n Fiqure 3-YQ. I n
t h i s case, since H (s) does not have a po le correspondinR t o t h e zero eq
1 of G ( s ) a t s = - - ' I ' 1
3
54
1. Figure 3-10. System i n which b and d a r e Related by a Constant, T
55
(3-28)
I n the above equat ions , -a is the pole of G ( s ) f o r which t h e r e is no
corresponding zero i n H (81, - - is t h e zero of G ( s ) f o r which t h e r e
is no corresponding pole i n H (s), and K' and K'r are t h e open loop
ga ins f o r t h e systems represented by Equations 3-27 and 3-28 r e spec t ive ly .
eq T1
eq 1
From Equations 3-27 and 3-28, it fol lows immediately t h a t t h e
frequency c r i t e r i o n f o r opt imal c o n t r o l is always s a t i s f i e d , s i n c e t h e
p o l a r p l o t of KG(e)H
t h e r e f o r e remains ou t s ide t h e un i t circle of Figure 3-9.
( 8 ) never c rosses i n t o t h e l e f t h a l f plane, and =q
From Figure 3-lc and Equation 3-27,
(3-29)
rJlsl = G(s), and D ( s ) has a zero a t s = -a. where
sen ted by Equation 3-28, t h i s becomes
For t h e case repre- D ( 8 )
(3-30) Y KN(S) fib) = q s + a + K K ' r l s + K K ' ] s+a
I n both these equat ions, .it is seen t h a t t h e (s+a) f a c t o r i n D ( s ) is
cancel led. It fol lows t h a t f o r a l l cases (n'l) of t h e c losed loop poles
are t h e same as (n-1) of t h e open loop poles .
w i l l have l i t t l e e f f e c t on t h e na tu re of t h e response i f K is made large
enough, s i n c e t h e r e s idue and time cons tan t a s soc ia t ed wi th i t become
n e g l i g i b l e as it moves f a r ou t from t h e o r i g i n .
The nth c losed loop po le
I f t h e state v a r i a b l e s are def ined in such a way t h a t t h e block
diagram of Figure 3-11 r e s u l t s , t h e expression f o r H
3-25 and 3-26 can be w r i t t e n in t h e same form as t h a t r e s u l t i n g from t h e
(s) i n Equations eq
56
Figure 3-11. System wi th n F i r s t Order Transfer Functions i n Series i n
the Forward Path.
57
block diagram formulation. Since b ' (and d ' i f Equation 3-26 a p p l i e s )
is non-zero only i n the nth element, H (5) becomes eq
(3-31)
a where F
t h a t cT = I 1
(s) is t h e nth column of Fa(s) . With t h e previous assumptions n
0 - 0 1 and t h a t kl = 1, Equation 3-31 reduces
f u r t h e r t o
a (s)+k2fn2 a (s)+"'+k f a (s) f n l n nn
Heq(s ) = fnl a (9)
a fnl (9)
a fnl (SI
( 3- 32)
a a where f ( 8 ) is t h e ith element of F (s). Since f o r t h i s conf igu ra t ion n i n
Equation 3-32 can b e w r i t t e n as
This has the same form as t h e expression f o r H ( s ) obtained from t h e
block diagram formulat ion by Schultz (1966). Thus t h e block diaRram
eq
formulat ion is seen t o represent a s p e c i a l case of t h e more general
mzt r ix formd.ation.
Some r e s t r i c t i o n s on the H ( 8 ) t h a t can be r e a l i z e d by feeding eq
back a l l t h e state v a r i a b l e s a re obvious. For example, i n Figure 3-11,
t h e zeros of H
G (s) s ince , as seen from Equation 3-33, t h i s would r e q u i r e t h a t kn = =.
Also, s i n c e t h e output is fed back d i r e c t l y t o t h e input , i t follows
t h a t H
a zero roo t . I n a system wi th one i n t e g r a t i o n , t h e zeros of H ( 8 ) are
forced t o equal t h e non-zero poles of G(s).
i n t e g r a t i o n s , H (s) cannot have (n-1) zeros equal t o (n-1) poles of G ( s ) .
( 8 ) cannot b e made equa l t o t h e poles of Gn - l ( s ) -m** eq
1
(s) must always have a constant term and t h e r e f o r e cannot have eq
eq I n a sys t emwf th two o r more
eq
On t h e b a s i s of t h e above d iscuss ion , t h e fol lowing procedure is
proposed f o r t h e design of gain i n s e n s i t i v e systems:
1.
2.
3.
4.
Describe t h e system i n terms of meaningful, phys ica l s ta te
v a r i a b l e s and assume t h a t t hese are a l l a v a i l a b l e and are
fed back through constant ga in elements.
Choose t h e des i red closed loop response $s).
Use a combination of series compensation and feedback t o
in su re t h a t a l l bu t one of t h e open loop poles correspond
t o the des i red closed loop poles . Normally, a pole a t t h e
o r i g i n is l e f t undisturbed so as not t o change t h e type of
t h e system.
Use state v a r i a b l e feedback t o fo rce t h e zeros of H
t o correspond t o the a l t e r e d open loop po le s , which are
t h e des i r ed closed loop poles .
ki can be found by c a l c u l a t i n g H
Y
( 8 ) eq
The requi red va lues of t h e
( 8 ) i n terms of t h e ki eq
59
by any of t h e above methods and equa t ing t h i s t o t h e
des i r ed H ( 8 ) . eq
5. I f a l l s t a t e va r i ab le s are not a v a i l a b l e , use t h e calcu-
l a t e d va lues of k t o determine s u i t a b l e minor loop com-
pensa t ion .
..
The fol lowing example i l l u s t r a t e s t h e des ign procedure.
Example 3-5: The system of Example 3-1 i s used. It is assumed t h a t t h e
d e s i r e d l o c a t i o n s of two of t h e closed-loop po le s are a t s - -2 f j 2 .
The o t h e r closed-loop pole is not s p e c i f i e d , b u t as shown above, moves
along the nega t ive- rea l axis as R is var ied . In Example 3-1, t h e numera-
t o r of G ( s ) w a s found t o b e 10. The denominator is r 1
O s+4 I + This g ives
10K s (sfl) (s+4) e
G(a) =
Some conf igu ra t ion f o r G(e) m u s t be assumed be fo re proceeding t o s t e p 3
of the design procedure.
t o f eed back x
a t t h e l o c a t i o n of t h e d e s i r e d c losed loop po le s of s = -2 2 j 2 .
This is shown i n F igure 3-12a. It is necessary
and x 2 3 as shown i n Figure 3-12bto have open loop poles
This
r e q u i r e s t h a t
-2
s+4+Klk s 2 +48+8 = d e t ri: 1
= s 2 +(5+Klk2')s+4+Klk2'+2K1.
60
r 5
s + 4
U , - K 7
J
1 Y = , x2 - x3 - 2
s + l S
A
Figure 3-12a. The P lan t to be Control led i n Example 3-5 .
4
1 y = x 2 1 - S
5 x3 2 x1 Y =
s + 4 s + 1 S K1 K U
* - J
I A
Figure 3-12b. Feedback Configurat ion Used t o Force the Open Loop Poles
t o be a t the Location of t he Desired Closed Loop Poles .
/ I - X - - 5 ~ 2 5
2 s + 4 s + l
Figure 3-12c. Method of Con t ro l l i ng the P lan t of Example 3-5.
k
61
Equating coe f f i c i ent s and solving the result ing equations gives
5 K1 = 2
2 k 2 1 = - 5 ’
Proceeding t o step 4 of the design procedure, the f i n a l feedback con-
f iguration required by S t e p 4 of the design procedure i s shown i n
Figure 3-12c.
S+l -2
- 5 s+3 i o I 2 -
( 3 - 3 4 )
Substituting into Equation 3-25 gives
Heq(s) = 2 kg [s(s+l)+ -
Equating t h i s t o the des ired value of H (s), eq
kg 2 H (9) = 7 [ s +4s+8], =I I
gives
- 4 . - 2k2
kg
Solving t h i s set of equations y i e l d s
1 k3 “z k2 = - . 1
2
62
Appl ica t ions of Gain I n s e n s i t i v e Design t o Reactor Cont ro l
I n t h i s s e c t i o n , t he ga in i n s e n s i t i v e design procedure ou t l ined i n
the previous s e c t i o n is app l i ed t o t h e c o n t r o l of t h e nuc lea r r e a c t o r I .-_- -
e-- , model wi th r e a c t i v i t y feedback due t o temperature , shown i n Fig. 3.13 where
p = i
p f =
n =
n = 0
x = A =
5" a -
intlut r e a c t i v i t y
feedback r e a c t i v i t y
r e a c t o r power
s teady s t a t e power l e v e l
weighted neutron p recu r so r decay cons t an t
neutron generat ion t i m e
cons t an t r e l a t i n g power t o r e a c t i v i t y feedback
h e a t removal c o e f f i c i e n t n
0 Observing Fig. 3.13 i t is noted t h a t t h e gain term, h, i s a func t ion of
t h e s t eady state power l e v e l and the re fo re w i l l vary as the power level
changes. This v a r i a t i o n w i l l change the po le l o c a t i o n s of t he c losed loop
t r a n s f e r func t ion , no and thus t h e system dvnamics. The design crite- Pi(S)
* r i o n is to determine a s ta te v a r i a b l e feedback-cont ro l such t h a t t h e system
dynamics is independent of power l eve l . For purpose of i l l u s t r a t i o n t h e
fol lowing va lues are assumed f o r the s y s t e m parameters
B A X = 0.1 - = 6 . 4 a = 2 KT = 5 x
It is f u r t h e r assumed t h a t t h e c o n t r o l l e r dynamics can be neglec ted i n
comparison t o the t i m e cons t an t s of t h e r e a c t o r . The assumption of a
p e r f e c t c o n t r o l l e r wi th a t r a n s f e r func t ion equal t o 1 is reasonable i n
space reactor systems the c o n t r o l l s r o have a verp fao t t i m e resp-.
6 3
4
,
I s + a I
Figure 3.13 Point Reactor with Reactivity Feedback
64
Applying s ta te v a r i a b l e
t o t h e r e a c t o r model i n Fig.
Fig. 3.14a, which reduces t o
feedback c o n t r o l , under t h e assumptions above,
3.13 gives a system of t h e form shown i n
t h e sys t em i n Fig. 3.14b, where
5x10-’ + kl + s+2 (3-35)
-9 If kl and k2 are l a r g e compared t o 10
neglec ted and Eq. (3-35) reduces t o
then the term involv ing lo-’ can be
(kl+k2) s+O . lkl
Heq(s) = s H . 1
L e t t i n g kl = 1 and assuming t h e des i red H ( 5 ) i s eq
(kl+k2) (s+6.4) Heq(s) = sw.1
(3-36)
(3-37)
then comparing Eqs. (3-36) and (3-37) gives
k 2 m - - 63 64
Clea r ly wi th these va lues f o r k
terms invo lv ing 10 is v a l i d . The system i n Fig. 3.14a now becomes of t h e
form shown i n Fig. 3.15.
and k2, t he assumption of neg lec t ing t h e 1 -9
The t r a n s f e r func t ion of the system i n Fig. 3.15, has po le s a t R ( s ) n
64 A 0 s = -6.4 and s - - - and a zero a t s ~-0.1. For va lues of - greater
n has very l i t t l e e f f e c t on system dynamics. For than 10 , t h e po le a t - -
64 A
a neutron genera t ion t i m e of A =
A 0 4
sec t h e corresponding power level
is 10 w a t t s . Therefore , f o r power r e a c t o r s , t h e system dynamics i s n
v i r t u a l l y independent of t h e gain and thus t h e r e a c t o r power l e v e l .
Since i t is impossible t o measure t h e state v a r i a b l e x2, t h e c o n t r o l is
r e a l i z e d by feeding back t h e r eac to r ou tput through a lead l a g network
0
- -._ .
v -- ~ - - _ _ -- c _
6 5
5~ IO-^ s+ 2 5~ IO-^ s+ 2
s+ 0.1 XI ‘ n o A s(s+ 6.4)
(b 1
Figure 3.14 Point Reactor with State Variable Feedback
66
r - n0 ‘/64 6 4 ( s + O . l ) s+6.4 A S
Figure 3.15 Equivalent Block Diagram for System i n Figure 3 .14
6 7
given by E q . (3-37) . The resultant design is shown i n Fig. 3 .16 . Un-
fortunately, the design of Fig. 3.16 may not be the transfer function
desired, even though i t i s independent of the reactor power l e v e l .
6 8
r + fJ (s + 0.1)
s( s + 6.4) A
Y
t- 0.016(s t 6.4) s+ 0.18
Figure 3.16 Final Reactor Control System Design
69
This is the d e s i r e d r e s u l t and, w i t h t h e assumption t h a t a l l t he s t a t e
v a r i a b l e s are a v a i l a b l e t o be fed back, completes t h e design procedure.
The closed loop t r a n s f e r funct ion is
Y 2 5K $S) = 2 25 (s + 4 ~ + 8 ) ( s + - 2 Kk3)
This has two po le s a t t h e l o c a t i o n of t h e modified open loop po le s and
a t h i r d po le a t s = - 2L\k3. 25 Thus i f K is l a r g e , t h e t h i r d p o l e w i l l
have l i t t l e e f f e c t on t h e system response, because of t h e f a s t t i m e con-
s t a n t and small r e s idue a s soc ia t ed w i t h it. The response is thus p r i -
mari ly dependent on t h e p a i r of complex po le s and independent of K.
Procedure When A l l S t a t e Variables Cannot B e Fed Back I
\ This t o p i c was discussed In jkmnples 3.4 and 3.5 and is
f u r t h e r emphasized In t h i s s e c t i o n by working an example.
3-1, i t i s assumed t h a t x2 cannot be f e d back.
determlned as before and x
I n example L .
The c o e f f i c i e n t s are
i s f e d back through a feedback func t ion k3 3 Y2(s) x- (9) Xi (SI
x3(s) u(s) '' can be found by notinR t h a t - can + k - This r a t i o of
2 X2(S)'
b e found i n the manner discussed p rev ious ly by l e t t i n g p = x i n Equation
3-2. This simply means t h a t ci i s t h e only non-zero element i n c . Per-
forming t h i s ope ra t ion i n t h e problem under cons ide ra t ion gives - =
i
x2 (SI x 3 ( s )
k3(~+1)+2k2 . This 2 - s+l Therefore, i s f ed back through k3 + k2 (s+l) - s+l x3
feedback func t ion can be r e a l i z e d wi th a pass ive network. It i s nex t
assumed t h a t only t h e output can be f e d back i n Example 3-1. Again, t h e
feedback c o e f f i c i e n t s are determined as be fo re , and x4 is de f ined as shown
i n Figure 3-7a and fed back through H1(s), where
70
Y =
Figure 3-7a.
be Fed Back.
Feedback Configuration When Only the Output Variable Can
Figure 3-7b. An Equivalent System Using Series Compensation. K
71
H1(s) = k2
From t h e v a l u e s of
H1(s) = k2
the example, t h i s is found t o be
10 5 10k2+5k3(s+1)
(s+l)(s+4) + k3 9+4 = (s+l)(s+4)
This feedback funct ion can a l s o be r e a l i z e d with a passive network. I n
f a c t , i t can be combined with K t o o b t a i n a series compensation network
as shown i n Figure 3-7b. This configurat ion i s similar t o t h a t obtained
by t h e Guillemin method,
On t h e b a s i s o f t h e above d i scuss ion , t h e Procedure t o be followed
when one o r more s ta te v a r i a b l e s cannot be fed back i s t o c a l c u l a t e t h e
feedback c o e f f i c i e n t s as though a l l s ta te v a r i a b l e s could be f ed back and
t o use t h e r e s u l t s i n determining p h y s i c a l l y r e a l i z a b l e feedback funct ions.
This is n o t t h e same as feeding back a l l t he s t a t e v a r i a b l e s f o r two rea-
sons: l) It assumes t h a t t he t r a n s f e r funct ion between t h e two states
is known e x a c t l y and can be reproduced exac t ly . This i s never t r u e .
Changes i n the system parameters would a f f e c t t h e two feedback confie;ura-
t i o n s d i f f e r e n t l y .
2)
Summary
A method f o r designing l i n e a r systems f o r a d e s i r e d closed loop
response by feeding back a l l the s ta te v a r i a b l e s has been developed from
t h e matrix r e p r e s e n t a t i o n of such systems. The procedure is s t r a i g h t -
forward and r e q u i r e s only elementary ma t r ix ope ra t ions and t h e s o l u t i o n
of n l i n e a r a l g e b r a i c equations. It p a r a l l e l s t h e procedure of Schul tz
(1966) which i s based on t h e block diagram system rep resen ta t ion . It i s
72
used t o develop a procedure f o r designing gain i n s e n s i t i v e svstems which
l e a d s t o t h e design method €or c e r t a i n nonl inear and/or time-varying
systems developed i n t h e following chapter .
Although the des ign procedure is r e f e r r e d t o as the SVF method,
i t inco rpora t e s t he c l a s s i c a l techniaue of series compensation wi th t h a t
of feeding back a l l t he state va r i ab le s . The use of s e r i e s compensation
makes i t p o s s i b l e t o add poles and zeros, thereby i n c r e a s i n g the o rde r
of t he system and inc reas ing the f l e x i b i l i t y of t h e design method. It
does n o t make use of t he information provided by the state v a r i a b l e s .
The use of s t a t e v a r i a b l e feedback does make use of t he information pro-
vided by a l l t h e state va r i ab le s , and t h i s in format ion is f ed back through
cons tan t ga in elements as suggested by t h e resu l t s of modem c o n t r o l theory.
An i n t e r e s t i n g c h a r a c t e r i s t i c of t he gain i n s e n s i t i v e systems desinned by
t h e method proposed he re is t ha t they always s a t i s f y the Kalman frequency
condi t ion f o r op t ima l i ty . H o s t systems designed by c l a s s i c a l techniques
do not s a t i s f y t h i s c r i t e r i o n .
CHAPTER I V
DESIGN OF NONLINEAR AND/OR TIME-VARYING CONTROL
SYSTEMS V I A STATE VARIABLE FEEDBACK
In t roduc t ion
I n t h i s chapter , a proposed method of s y n t h e s i s f o r s ing le- input ,
s ingle-output systems containing a s i n g l e nonl inear and/or time-varying
ga in which s a t i s f i e s condi t ions 2-4 ( o r 2-5) and 2-6 i s developed. This
is accomplished by compensating the cons tan t l i n e a r po r t ion of t h e system
i n such a way t h a t t h e Popov s t a b i l i t y c r i t e r i o n discussed i n Chapter 2
is s a t i s f i e d . The design procedure i s a l o g i c a l ex tens ion of t he state
v a r i a b l e feedback design of l i n e a r ga in i n s e n s i t i v e systems as developed
i n Chapter 3.
an i n f i n i t e s t a b i l i t y s e c t o r and t o have a bounded output f o r bounded
Systems designed by t h e proposed method are shown t o have
inputs .
The degree of success w i t h which t h e method can be used i s dependent
upon t h e form of t h e p a r t i c u l a r system.
c h a r a c t e r i s t i c of a n a l y s i s and design procedures f o r nonl inear systems.
I n i t s b a s i c form, t h e method is l imi t ed t o systems wi th one nonl inear and/
o r time-varying ga in loca ted as shown i n Figure 4-1.
systems having no more than one i n t e g r a t i o n noted i n t h e d i scuss ion of
l inear ga in i n s e n s i t i v e systems a p p l i e s here a l so . It is shown i n the
next chapter t h a t t h e method i s no t app l i cab le t o p a r t i c u l a r cases o t h e r
than t h e s imples t p a r t i c u l a r case due t o s t r u c t u r a l s t a b i l i t y problems.
This l a c k of g e n e r a l i t y is
The l i m i t a t i o n t o
73
74
Thus the method i n i t s bas i c fonn i s app l i cab le only f o r t he p r i n c i p a l
case and the s imples t p a r t i c u l a r case . Modif ica t ions i n the design
procedure which remove some of these r e s t r i c t i o n s i n c e r t a i n cases a r e
considered i n Chapter 5.
The proposed method is app l i cab le , i n a p r a c t i c a l sense, t o
systems of any order . The determinat ion of t he feedback c o e f f i c i e n t s
r e q u i r e s the s o l u t i o n of (n-1) l i n e a r a l g e b r a i c equat ions , where n i s
the o rde r of t h e l i n e a r system G ( s ) . formula t ions of the procedure a re d iscussed .
Both the matrix and block diagram
The organiza t ion of t h e remainder of t h i s chap te r i s as fol lows:
1) For systems con ta in ing nonl inear ga ins , t he bas i c design procedure
i s developed, and t he abso lu te s t a b i l i t y p r o p e r t i e s and the c losed loop
response of the r e s u l t i n g system a r e d iscussed . 2) These same t h r e e
t o p i c s are d iscussed w i t h respec t t o t ime-varying ( o r nonl inear and t i m e -
varying) systems. 3) The s i g n i f i c a n t f e a t u r e s of systems designed by
t h e proposed method a r e summarized.
The SVF Method f o r Nonlinear Systems
Figure 4-1 i l l u s t r a t e s t he b a s i c feedback conf igu ra t ion of t h e
compensated system.
series w i t h a s t a b l e l i n e a r system, G ( s ) .
i n t h e p l a n t t o be c o n t r o l l e d as an undes i rab le c h a r a c t e r i s t i c , o r i t
may be i n t e n t i o n a l l y introduced i n o rde r t o achieve a des i r ed r e s u l t .
For example, a s a t u r a t i o n element might be used t o prevent signals i n
some p a r t of the system from becoming excess ive .
The system c o n s i s t s of a s i n g l e nonl inear ga in , N,in
The nonl inear gain may appear
The system equat ions
a r e given i n Chapter 2 and a r e repea ted here .
75
Figure 4-1.
a Single Nonlinear and/or Time-Varying Gain.
Basic Configuration f o r Con t ro l l i ng a Plant Containing
u = f ( a )
76
( 4 - 1 )
( 4 - 2 )
( 4 - 3 )
( 4 - 4 )
A s i n t h e case of t h e l i n e a r systems of Chapter 3 , s ince the state
v a r i a b l e s are r equ i r ed t o correspond t o physical v a r i a b l e s , Equation 4-1
might have the form
2 .. = Ax+bu+dG. ( 4 - 5 ) -.. - - The matr ix formulation of t he b a s i c design procedure i s e x a c t l y
the same a s t h a t developed f o r l i n e a r gain i n s e n s i t i v e systems i n the
previous chap te r and i s the re fo re n o t repeated here . It i s noted t h a t the
ga in K appears nowhere i n t h e t r a n s f e r func t ions used i n the c a l c u l a t i o n
of t h e ki f o r gain i n s e n s i t i v e systems and t h e r e f o r e does not a f f e c t
H e q ( s ) . Thus t h e f a c t t h a t K i s now assumed t o be nonl inear has no e f f e c t
on t h e procedure f o r f ind ing II (s) . The only equat ions i n the d i scuss ion
of Chapter 3 which cannot be appl ied i n the nonl inear case are those f o r
eq
t h e c losed loop t r a n s f e r funct ion. The r e fe rence t o c losed loop poles i n
t h e procedure i s j u s t i f i e d f o r the nonl inear case i n t h e s e c t i o n on c losed
loop response.
The block diagram formulation of t h e design procedure r e q u i r e s
t h a t t h e block diagram be manipulated i n t o the series form shown i n
Figure 4-2a, where t h e G ( s ) are f i r s t o rde r t r a n s f e r func t ions . The
feedback conf igu ra t ion i s shown i n Figure 4-2b, and H (s) can e a s i l y be eq
determined by comparing t h i s with the equ iva len t system of Figure 4-2c.
The r e s u l t i s
i
77
Figure 4-2a. Block Diagram i n t h e Form of a S e r i e s of F i r s t Order
Transfer Funct ions.
Figure 4-2b. Method of Contro l l ing the System of Figure 4-2a.
Figure 4-2c. An Equivalent System f o r F igure 4 - 2 b .
Heq(s) = l+k
which is the same as Equation 3-33 developed from
78 -1
( 4 - 6 ) 9
the matrix
representation for this particular configuration.
Absolute Stability of the Resulting System
The two possible open loop transfer functions for a system
designed by the proposed method are given in Equations 3-27 and 3-28
and are repeated here:
(3-27)
(3-2 8)
The modified plot used in the interpretation of the Popov criterion
never crosses into the left half plane for either of these functions.
Therefore, the Popov line can be drawn through the origin, indicating
that the sector i n xhich the system is absolutely stable for the type
nonlinearity being considered includes the entire first and third
quadrants.
Closed Loop Response of the Resulting System
A mathematically equivalent system is derived for studying
input-output relations of the closed loop system. The linear system
G ( s ) is rearranged as shown in Figure 4-3a, where G (s) is the first a
orde r t Tans f e r
order transfer
function of Equation 3-27 or 3-28, G ' ( s ) is an (n-1)st
function given by
P Go K1'Ga(s) ' (4-7)
79
N , f (a) G,(s) G ' ( s ) , Y =
Figure 4-3a.
with the Linear Part Rearranged,
The Equivalent System of Figure 4-1 Showing H (s) and eq
Figure 4-3b. The Equivalent System of Figures 4-3a and 4-1 for
Determining Input-Output Relations.
80
n- 1 and K" is th2 c o e f f i c i e n t of s i n H (5). Figure 4-3a is next
represented by t h e equ iva len t s y s t e m of Figure 4-3b, which is used i n
t h e a n a l y s i s t h a t fol lows. The followinR observa t ions are made con-
ce rn ing the equ iva len t system:
eq
1.
2.
3 .
4.
The state v a r i a b l e s x$,---, x: are d i f f e r e n t from
those of F igure 4-1. Only (J, f ( u ) , and t h e inpu t and
output v a r i a b l e s a r e the same.
Changes i n system parameters w i l l a f f e c t t he systems
of F igures 4-1 and 4-3 d i f f e r e n t l y .
The equ iva len t system c o n s i s t s of a f i r s t o rder nonl inear
system w i t h u n i t y feedback i n s e r i e s w i th an ( n - 1 ) s t o rder
s t a b l e l i n e a r system. Therefore , i f i t can be shown
t h a t the output of the nonl inear po r t ion is bounded f o r
a bounded inpu t , i t fo l lows t h a t t he output , xl, of t h e
a c t u a l system represented by Figure 4-1 i s a l s o bounded.
The equ iva len t system i n d i c a t e s t h a t t h e (n-1) po les of
G'(s) are a l s o poles of the c losed loop system.
The Second Method of Liapunov i s used t o show t h a t t h e output
i s bounded f o r a bounded inpu t .
has t h e form of Equation 3-28.
po r t ion of t he equiva len t system a r e then
For g e n e r a l i t y , it i s assumed t h a t G,(s)
The system equat ions f o r t he nonl inear
5' n = -ax'+K'f(a)+K'.r,k(u> n
u = r - X I . n
A Liapunov function of the form
2 v = x' n
is chosen. It follows that
ir - 2x; [ - a A + K ' f (a)+K'~~*(u) 1.
81
(4-10)
(4-11)
From Equation 4-9 and the nature of the nonlinearity as given by
conditions 2-4 through 2-6, it follows that
when r is bounded if the magnitude of 4 becomes large enough. also true for the cases where G , ( s ) has no zero (T
G , ( s ) has an integration (a = 0), or both.
linear portion of the equivalent system is bounded for all cases when the
input is bounded.
equivalent system is a stable linear system, it follows that the output
of the closed loop system of Figure 4-1 is bounded for bounded inputs.
Moreover, since the output of the nonlinearity is the same in the
equivalent system of Figure 4-3b and in the actual system of Figure 4-1,
all t he physical state variables are bounded, or the aysten: is Lzgrange
stable.
will always become negative
This is
= 0) and where
Thus the output of the non-
Since the system from x: to the output y of the
Under certain quite restrictive conditions, an analytical
s,)lution for the output of the nonlinear portion of the equivalent system
can be obtained. A sufficient condition for obtaining such a solution is
that the variables in Equation 4-8 can be separated and the resulting
integration carried out. This can be accomplished in the simplest
particular case (a = 0) withT1 = 0 if the input is a step function and
f(u) is such that
the step input. The following example illustrates the procedure.
dx can be integrated, where M is the magnitude of f (M-IC)
82 3 Example 4-1: It is assumed that r = Mu(t), f(a) = o , a = 0, and = 0.
Then Equation 4-8 may be written in the form
1 1
-
t f
' I - 2 .
(4-12)
Carrying out this integration gives
= t. 1 - 1 ~(M-x')~ 2(M-x;(0) 2
n Rearranging and solving for < gives the result
i- - 1
(4-13)
As t becomes large, x:
this particular input.
* M, the equilibrium state of the system for
The conditions necessary for obtaining an analytical solution
for x' are so restrictive that some other means of determining this output
of the nonlinear part of the equivalent system is desirable. Since the
nonlinear part is a first order system, it is always possible to find
n
the output by use of the basic graphical procedure known as the isocline
method when the input and the nature of the nonlinearity are known. This
is illustrated by the following example.
1 Example 4-2:
type gain described by Figure 4-4a.
It is assumed that Ga(s) = -and that f(a) is a saturation S
The nonlinear portion of the
equivalent system is shown in Figure 4-4b.
needed since this example is concerned only with determining the response
This is all the information
/I
Figure 4-4a. Nonlinear Gain Characteristic for Example 4-2.
-I I I
Figure 4-4b. Nonlinear Portion of the Equivalent System for Example
4-2.
84
of t h e nonl inear p o r t i o n of t he equiva len t system. The i s o c l i n e method
i s used t o determine t h i s response f o r a s i n u s o i d a l i n p u t , r = n s i n t.
From Figure 4-4b,
(I = TI s i n t-x' (4-14a) n'
ic' = f ( a ) . (4-14b) n
From t h e s e r e l a t i o n s h i p s and the nonl inear c h a r a c t e r i s t i c s shown i n
Figure 4-4a, t he fol lowing va lues of 5' a r e determined: n
1. When x; > 1 + n s i n t , ic; = -1.
2. \ h e n x; < -1 + TI s i n t , k' = 1. n
3. When x; = h + 71 s i n t f o r -1 < h 1, 5; = -h. - -
These va lues of 5' are used t o determine x' g raph ica l ly a s shown i n n n
Figure 4-5.
equ iva len t system, and the output can now be found by l i n e a r methods.
This response i s the inpu t t o the l i n e a r po r t ion of t he
The a b i l i t y t o show t h a t t h e closed loop system has a bounded
output f o r bounded inpu t s and t o determine t h e output f o r a s p e c i f i c
i npu t i s a s i g n i f i c a n t advantage, as t h i s is no t gene ra l ly Dossibie wi th
non l inea r systems. The next example , - i l l u s t r a t e s t he a p p l i c a t i o n of t h e
b a s i c design procedure and the c a l c u l a t i o n o f t h e c losed loop response
of t he r e s u l t i n g system.
Example 4-3:
i n Figure 4-6a. This system might r ep resen t , f o r example, a DC motor
dr iven by an ampl i f i e r which s a t u r a t e s for large inpu t s . I n t h i s case
x r e p r e s e n t s p o s i t i o n , x2 ve loc i ty , and x t he f i e l d cu r ren t . The
s a t u r a t i o n l e v e l might be inherent i n t h e a m p l i f i e r o r i t might be b u i l t
The block diagram of the system t o be c o n t r o l l e d is shown
1 3
85
e aJ U (0
h a
C aJ
z U 7 0
86
~
5 1 x2 L - x3 - . N s + 5 s + l ' s
u= f(0) U
r - -
4
Y =
i
Figure 4-6a. System t o be Control led i n Example 4-3.
I I - J
5 - - , K1 s + 5 N U
_1
Figure 4-6b. Compensation of the System t o Force t h e Open Loop Poles t o
be Equal t o the Desired Closed Loop P o l e s ,
I U
Figure 4-6c. Method of Con t ro l l i ng the Plan t of Example 4-3.
Figure 4-6d. Equivalent System f o r F igure 4-5c.
87
i n t o the a m p l i f i e r t o prevent the v e l o c i t y from exceeding some maximum
va lue .
i n a s a t e l l i t e t r a c k i n g system.
system can be very l a r g e dur ing the process of l o c a t i n g the t a r g e t , t hus
d r i v i n g the a m p l i f i e r wel l i n t o the s a t u r a t i o n region. Once the t a r g e t
i s loca ted and being t racked , la1 i s assumed t o be s m a l l enough t h a t t he
An a p p l i c a t i o n f o r such a system might be t o p o s i t i o n an antenna
The a c t i v a t i n g s i g n a l , a , i n such a
system ope ra t e s i n the l i n e a r region.
f o r any a c t i v a t i n g s i g n a l and should respond quickly f o r small a c t i v a t i n g
Thus t h e system must be s t a b l e
s i g n a l s .
t h e ga in of t he a m p l i f i e r i s high i n t h e l i n e a r region and that the
I n order t o achieve the d e s i r e d response, i t i s assumed t h a t
c losed loop system has poles a t s = - 2 - + j 3 . The f i r s t s t e p i n the
des ign procedure i s t o compensate the system as shown i n F igure 4-6b
so t h a t the open loop poles a r e a t t he des i r ed l o c a t i o n of t he c losed
loop poles .
-7 and K1 = 2 r e s u l t i n t he open loop poles having the desired l o c a t i o n s .
The s t a t e v a r i a b l e s a r e then f e d back a s i nd ica t ed i n F igure 4-6c.
By t h e method of Chapter 3 , i t i s determined t h a t k i = 1
I n
the equ iva len t system nf Figure 4-6d,
2 k2 1
k3 k3 Heq(s) = k3S(S+1) + k 2 ~ + 1 = k3 [ s + ( l+- )s + --I. (4-15)
I n o rde r t o r e a l i z e t h e des i red c losed loop-poles , t he requi red Heq(s)
must have zeros a t s = -2 - + j 3 . This gives . 2 H (s) = k3 [s + 4s + 13 eq
Equating the c o e f f i c i e n t s of Equations 4-15 and 4-16 g ives
k2
kg 1 + - = 4
(4-16)
Thus the requi red feedback c o e f f i c i e n t s a r e
1 k g =,E k2 a - 0 3
13 It has-already been shown t h a t t h e system when designed i n the
above manner w i l l be a b s o l u t e l y s t a b l e r ega rd le s s of the gain i n the
l i n e a r region of ope ra t ion . I n add i t ion , the l o c a t i o n of t he two
complex po le s is independent of the gain. Therefore , the ga in can be
made as h igh as des i r ed i n the l i n e a r region, and t h e r e s u l t i n g system
h a s the d e s i r a b l e c h a r a c t e r i s t i c s of s t a b i l i t y f o r any e r r o r s i g n a l and
the des i r ed response i n the l i n e a r reg ion of ope ra t ion ,
While the sys t em is i n the l i n e a r t r ack ing mode, t he c losed loop
t r a n s f e r func t ion i s
Y 10K $ 8 )
( 8 +4s+13)+ E K(s2+4s+13)
10K 9 (4-17) = 2 10 (s +4s+13) (s+ 5 K)
where K i s the ampl i f i e r ga in i n t h e l i n e a r region. This i n d i c a t e s t h a t
t he c losed loop system has poles a t s Both the 10
-2 c f j 3 and s a -5 K. t i m e cons t an t and the r e s idue a s soc ia t ed wi th t h e real pole are very
small, and the na tu re of t he system re sponse - i s t h e r e f o r e determined by
the p a i r of complex conjugate poles i f K i s l a r g e . The v e l o c i t y e r r o r
c o e f f i c i e n t is a l s o determined to a good approximation by the complex
conjugate po le s i f t h e ga in i n the l i nea r region is high. This i s given
by (Truxal , 1955)
3 (4-18)
I I 1 I I I 1 ,I I I I 1 I I i 1 1 I 1
89
This indicates that as K becomes large, K is determined by the location
of the complex conjugate poles. The significance of these results is
that the system can be designed for the desired characteristics in the
linear region without being bothered by stability problems when operating
in the nonlinear region.
V
The equivalent system for studying the closed loop response is
shown in Figure 4-7a. For the purpose of illustrating the procedure
for calculating the response from this equivalent system, it is assumed
that the amplifier has the characteristics shown in Figure 4-7b. It is
also assumed that at t = 0, the antenna is pointing in a direction 10
ahead of the satellite in the line of travel and that the satellite is
moving with respect to the antenna at the rate of one degree per minute.
The location of the satellite is taken as the reference position. Thus
the system begins operating in the saturation region, with an activating
signal of -10 . alent system is found by the isocline method and is shown in Figure 4-8.
ine relerence signa?, r = t, i s also shown inthis figure. When
-1 + t c x i c 1 + t, l a l < 1 and the system is operating in the linear regiono The graph shows chat the system operates in the linear region
after approximately 0.4 seconds. With the xi of Figure 4-8 as the input
to the linear portion of the equivalent system, the output of the closed
loop system can be found. One way of calculating this output is to
approximate the input to the linear portion of the equivalent system by
a piece-wise linear function. Due to the low-pass characteristics of
most control systems, this approach will usually yield a good approxi-
mation to the output. Where this approach is not practical, graphical
convolution can be used.
0
0 The output, x i , of the nonlinear portion of the equiv-
d
- - -
I I . 1 I c I 1 I I I I I I I 1 i I 1 I
, fP> - 10
13s N
90
1 X’ 13 y = x 3 ’ sL+4s + 13 ’
r d
Figure 4 - 7 a .
in Example 4 - 3 .
Equivalent System for Determining the Closed Loop Response
, b
30
Figure 4 - 7 b . Nonlinear Amplifier Characteristics for Example 4 - 3 .
I . I I 3 1 I I I I I I 1 I I I ~1 ,I I
4 -*
0 a3 4
91
L1
t
4
m
N
' 4 L N 0 N
I
92
In this example, x;, can be approximated by the function
x;(t) 8 10u(t)-23tu(tj i 24(t-.42)u(t-.42). (4-14 j
(It is nored that u(t) represents a unit step function and is not the
same u as is used for the system c o n t r o l function thrniighout t h i s
dissertation.) From Equation 4-19,
(4-20)
Combining this expression with the transfer function of the linear
portion of the system and noting that both x;(t) and y(t) are equal to
l o o at t = o gives, after rearranging, -. 42s 23 17 1 5.4+7.1s + 24e 2 Y ( s ) = - - + & -
s2+4s+1 3 S 2 8
S
-. 42s 7.4e-*42s 5.6+7.4s) e
s +4s+13 + ( 2 -
0 9
(4-21)
From this equation, after combining terms and simplifying,
y(t) f 17.lu(t)-7.4u(t-.42)-23tu( t)+24( t-. 42)u(t-. 42)
- 7; hSe-2tcos(3t+22. S O ) U ( t >
+ 8. 26e-2(t-*42)c0s [ 3 ( t - . 42)+22.5O]u(t-. 42) . (4-22)
Equation 4-22 is the approximate output of the system operating under
the assumed conditions. The graph of this output is shown in Figure 4-9.
The system of Example 4 - 3 was simulated on an analog computer,
and the resulting response shown in Figure 4-10 agrees reasonably well
with the calculated response shown in Figure 4-9. The gain in the
linear region was varied between 12 and 700 with the only noticeable
change in the response being a small increase in the overshoot as the gain
was increased. Thus the results of the simulation indicate that the
system is indeed insensitive to large variations in the gain.
I I I I i I I I 1 I I I 1 1 1 I I 8 I
I
\ \
X” II
h
Xd U h
0 rl
9 3
u
t h
rD
In
4.
c)
r\l
a ‘ 4
Y I I
I I
0 a3 F 4
n
h 5 -
9 hl
94
0
i
-- ?
:
.rl U rj rl
cn bo 0 t-l
a, 5 $4 w a a, C .rl (d U D 0
(d
m I .j
a, rl a E c5 X Lil
W 0
E a, U rn h v)
a,
rn
5 W 0 U 1 a U 1 0
0 rl
I U
Q)
3 M .d r=
95
The SVF Method for Time-Varying Systems
In this section, the application of the proposed design procedure
to time-varying (or time varying and nonlinear) gains is considered.
The system equations are the same as in the nonlinear case except that
Equation 4-2 becomes
The design procedure for time-varying systems is exactly the same
as that for nonlinear systems. Equations 3-27 and 3-28 indicate that
the modified plot of G ( s ) H ( s ) for the resulting system never crosses
into the second or third quadrants. Therefore the Popov criterion for
time-varying systems as given in Equation 2-17 is always satisfied, and
eq
the system is absolutely stable for all time-varying (or time-varying
and nonlinear) gains of the type being considered. An equivalent system
for input-output relations can be derived in the same manner as for non-
linear systems, with the only difference being that the gain in the system
of Figure 4-3b will be time-varying. Because of the restrictions on the
time varying gain as given by conditions 2-4 through 2-6, Equation 4-10
can again be used as a V function to show that the output is bounded for
bounded inputs.
The system configuration of Example 4-3 might also be used as
an illustration of the application of the design procedure to time-
varying systems. In fact, the feedback coefficients calculated in this
example would be the same for any nonlinear and/or time varying gain
(with the exception noted below) as long as the desired location of the
closed loop poles remain constant. The only exception is that the gain
cannot be equal to zero for u # 0. This restriction follows from the
96
Summary
me method of designing l i n e a r gain i n s e n s i t i v e systems has
been extended t o develop a design procedure f o r a c e r t a i n class of non-
l i n e a r and/or time-varying systems. The procedure i s based on r e a l i z i n g
a desired. equivaient feedback c r a n s f e r func t ion , neq(s j , by feeding back
a l l the s t a t e v a r i a b l e s i n t h e proper l i n e a r combination. The r e s u l t i n g
system has the following s i g n i f i c a n t f e a t u r e s :
1. The abso lu te s t a b i l i t y s e c t o r , as determined from t h e Popov
theorem, inc ludes the e n t i r e f i r s t and t h i r d quadrants.
2 . The output of the closed loop system i s bounded f o r bounded
i n p u t s . 3 . The c losed loop system can be represented by an equivalent
system f o r input-output r e l a t i o n s c o n s i s t i n g of a f i r s t o rde r
non l inea r and/or t i m e varying po r t ion i n series w i t h an ( n - 1 ) s t
o r d e r s t a b l e l i n e a r po r t ion . This equ iva len t system can be
used t o f i n d the approximate output f o r a given inpu t .
4 . i n c e r t a i n cases , the dominant t i m e constanc, band widch,
and overshoot of t h e c losed loop system f o r any input are
determined pr imari ly by t h e cons t an t l i n e a r po r t ion of t he
equ iva len t system and are e s s e n t i a l l y independent of t h e non-
l i n e a r and/or t i m e va ry ing gain.
I n systems where the ope ra t ion i s l i n e a r f o r normal c o n t r o l
s i g n a l s but nonl inear f o r l a r g e c o n t r o l s i g n a l s , i t i s poss ib l e
t o design f o r t h e desired ope ra t ion i n the l i n e a r region
without having t o worry about system s t a b i l i t y f o r l a r g e
c o n t r o l s i g n a l s .
5 .
97
6 . The closed loop system i s l i n e a r i z e d t o the e x t e n t t h a t
(n-1) of the closed loop poles a r e equal t o (n-1) of t he
open loop po le s of the l i n e a r p l a n t , independent of t he non-
l i n e a r gain.
The design procedure, unlike most non l inea r and/or time-
va ry ing design methods, presents no real computational d i f f i c u l t i e s .
It r e q u i r e s the s o l u t i o n of n l i n e a r a lgeb ra i c equat ions where n i s the
o r d e r of the l i n e a r p a r t of the system, G ( s ) . It i s t h e r e f o r e a p p l i c a b l e ,
i n a p r a c t i c a l sense, t o high order systems.
I n general , series compensation cannot be used i n the bas i c
des ign procedure un le s s such compensation can be loca ted i n t h e system
so t h a t t h e non l inea r and/or t i m e varying gain remains i n the r e l a t i v e
p o s i t i o n i n the system shown i n Figure 4-1. I n the next chap te r , consid-
e r a t i o n i s given t o some p r a c t i c a l a s p e c t s of t h e design procedure and
t o i t s poss ib l e extension t o include systems t o which t h e bas i c procedure
cannot be app l i ed .
CHAPTER V
I n t r o d u c t i o n
The design procedure of t h e previous chap te r is based on the
assumption t h a t a l l t h e s t a t e v a r i a b l e s can be measured and fed back
i n t h e proper l i n e a r combination t o f o r c e t h e zeros of t h e r e s u l t i n g
H ( 8 ) t o be e x a c t l y equal t o n-1 of t h e poles of G(s) . Since exac t
pole-zero c a n c e l l a t i o n i s never p o s s i b l e i n a phys ica l system, the
e f f e c t of small d i f f e r e n c e s between t h e , z e r o l o c a t i o n s of H ( s ) and
t h e corresponding po le l o c a t i o n s of G(s) i s of i n t e r e s t .
is considered i n t h e s e c t i o n on s t r u c t u r a l s t a b i l i t y a t t h e beginnine
of t h i s chapter .
eq
eq This ques t ion
The l a s t p a r t of t h i s chapter i s concerned with p o s s i b l e methods
of extending t h e design procedure t o c e r t a i n svstems where t h e b a s i c
procedure cannot be used. F i r s t , p o s s i b l e methods by which t h e b a s i c
procedure can be extended to sys t ems i n which t h e non l inea r and/or
time-varying gain i s no t located i n t h e p o s i t i o n shown I n Figure 4-1
are discussed. Then t h e p o s s i b i l i t y of using t h e design procedure t o
design systems having a f i n i t e s t a b i l i t y s e c t o r i n c e r t a i n cases where
an i n f i n i t e s e c t o r of s t a b i l i t y cannot be achieved is considered.
a r educ t ion i n t h e s e c t o r of s t a b i l i t y r e s u l t s i n less s t r i n g e n t con-
s t r a i n t s on t h e open loop Rain, C ( s ) H ( 8 ) .
Such
eq
98
99
S t r u c t u r a l S t a b i l i t y of t he System
The d i scuss ion of s t a b i l i t y i n t h e previous chapter i s based on
t h e p r e m i s e t h a t t h e zeros of H (s) can be made t o occur a t the exact
i ocac ions or' the poies of G(a i . Since percrct p o l e z e r o eziics1latiori
cannot be achieved i n a physical system, t h e r e w i l l be d i f f e r e n c e s i n
t h e l o c a t i o n s of t h e zeros of H (s) and t h e correspondina poles of G(s) .
A s a consequence, t h e expressions f o r G ( s ) H (s) Riven i n Equations 3-27
and 3-28 w i l l have n-1 a d d i t i o n a l po le s and zeros. These a d d i t i o n a l
c r i t i ca l frequencies have t h e property t h a t t h e zeros are almost equal
t o t h e poles . Thus t h e question of i n t e r e s t becomes: "How u s e f u l is
t h e approximation of Equations 3-27 and 3-28 i n a phys ica l system?"
answer t o t h i s quest ion l i es in t h e s t r u c t u r a l s t a b i l i t y p r o p e r t i e s of t h e
system. I f i t can b e shown t h a t small chanRes i n t h e system parameters do
n o t r a d i c a l l y a f f e c t t h e nature of t h e system response, then r e s u l t s ob-
t a i n e d from Equations 3-27 and 3-28 should be v a l i d approximations t o t h e
eq
eq
eq
The
a c t u a l system respoiise.
The e f f e c t of small d i f f e rences i n t h e corresponding po le s and
ze ros of G(s) and H (8) on the abso lu te s t a b i l i t y of t h e system i s con-
s i d e r e d f i r s t . Using t h e Popov s t a b i l i t y c r i t e r i o n , a n a l y t i c a l r e s u l t s
are obtained which i n d i c a t e t h a t t h e abso lu te s t a b i l i t y is no t g r e a t l y
a f f e c t e d by small changes i n system parameters except i n those cases where
closed loop po le s occur on o r nea r t h e imaginary axis.
i n d i c a t e how much v a r i a t i o n i n t h e po le and ze ro l o c a t i o n s can be t o l e r -
a t e d without t h e abso lu te s t a b i l i t y p r o p e r t i e s of t h e system beinn
s e r i o u s l y a f f ec t ed .
eq
These r e s u l t s
100
The quest ion of absolute s t a b i l i t y i s i n v e s t i q a t e d from t h e stand-
p o i n t of t he a d d i t i o n a l phase s h i f t introduced i n t o C ( s ) H
d i f f e r e n c e s i n corresponding pole and zero loca t ions . Zeros of H (s)
and t h e corresponding poles of G(s) t o which they are n e a r l y equal are
r e f e r r e d t o as pole-zero p a i r s i n t h e d i scuss ion which follows.
( 8 ) by small eq
eq
Only t h e e f f e c t of a s i n g l e pole-zero p a i r i s considered for po les
and zeros on t h e real a x i s . For complex conjugate po le s , two pole-zero
p a i r s must be considered. The r e s u l t s can then be used t o determine t h e
e f f e c t of d i f f e r e n c e s i n t h e zero and po le l o c a t i o n s of any pole-zero p a i r
i n t h e system.
The diagram of Figure 5-1 is used t o determine Reneral r e s u l t s
f o r a r b i t r a r y pole-zero loca t ions which are then used i n d i scuss ing more
s p e c i f i c cases. From Figure 5-1 t h e t o t a l phase s h i f t of t h e two p a i r s
of complex conjugate po le s and zeros i s
Tie first two terms rep resen t the phase s h i f t con t r ibu ted by t h e poie-
zero p a i r i n t h e upper h a l f plane, and t h e las t two terms rep resen t t h e
phase s h i f t con t r ibu ted by t h e lower h a l f plane pole-zero p a i r .
t h e expressions f o r t he phase s h i f t con t r ibu ted bv t h e two pole-zero pairs
are similar, t h e l a s t two terms are nefqlected f o r t h e p re sen t , piving
Since
From t h i s ,
a b 2 2 2 = E (a-b)-2acg+a(b +c -ab). - = - -
2 2 d E a2+Q2 b +(k-c)
( 5 - 3 )
S e t t i n g Equation 5-3 equal t o zero gives the fo1lowinr;t expression f o r t h e
values of I a t which 6 reaches its extreme p o s i t i v e and nega t ive values: 1
10 1
Figure 5-1.
the Pole-Zero P a i r s .
Diagram Used i n Determining the Phase S h i f t Contributed by
ac +'ab(a-b) 2 +abc 2 - R = 0
a-b
102
( 5 - 4 )
_- The maximum phase s h i f t due t o the complex pole-zero p a i r can be found by
s u b s t i t u t i n g the va lue of 11 found from Equation 5-4 i n t o Equation 5-2.
maximum to ta l phase s h i f t cont r ibu ted by the two complex coniuaa te gole-
zero p a i r s w i l l always be l e s s than twice t h i s value.
Equations 5-4 and 5-2 can a l s o be used t o determine the maximum phase
The
I
With C 0,
s h i f t due t o a pole-zero p a i r on the real a x i s .
Equations 5-2 through 5-4 a r e now used t o i n v e s t i a a t e s p e c i f i c
cases of po le and zero loca t ions .
Case I: The pole and zero are real wi th a = 0.9b. Riving a 109 d i f f e r e n c e
between t h e pole and zero loca t ions . From Equation 5-4 t he maximum phase
s h i f t occurs a t
-
R = 5 Jo,sb2 - 2 0.95b.
Since only p o s i t i v e va lues of frequency are of interest, only the nega t ive
va lue of is considered. S u b s t i t u t i n g i n t o Equation 5-2 gives
( 5 - 5 ) -1 0 = - tan (1.06) + tan-'(0.95) = -3.2 . maX
e
For b = 0.9a, t h e maximum phase s h i f t has t h e same magnitude b u t oppos i te
s ign . Thus t h e maximum phase s h i f t for a 10% d i f f e r e n c e i n t h e l o c a t i o n s
of a real axis pole and ze ro is 5 3.2'.
Case 11: There are two complex conjugate pole-zero p a i r s wi th c = 0 and
the r a t i o between a and b = 0.9. Here, the magnitude of t h e t o t a l phase
s h i f t from t h e two pole-zero p a i r s i s always less than twice t h a t of
Equation 5-5, o r 6.4'. This is a conserva t ive upper l i m i t f o r t he phase
103
s h i f t , as i t w i l l i n gene ra l be considerably less, s i n c e t h e maximum
from t h e two pole-zero p a i r s w i l l no t occur a t t h e same frequency.
Case 111: There are two complex conjugate pole-zero p a i r s wi th a = 0.9b
and c = O.lb. From Equation 5-4, t h e maximum phase s h i f t caused by t h i s
d i f f e r e n c e i n t h e upper h a l f plane pole-zero p a i r is found t o occur a t
R = 0.44b.
s h i f t of 7.2'.
t h e lower h a l f p l ane is no more than t h i s , t h e maximum phase s h i f t con-
t r i b u t e d by the complex conjugate pole-zero p a i r s i s seen t o be less than
14.4'. Again t h i s is a conserva t ive upper l i m i t f o r t h e maximum phase
s h i f t , which w i l l i n genera l be considerably less.
Case I V :
t o t a l phase s h i f t f o r p o s i t i v e f requencies comes from the pole-zero p a i r
i n t h e upper h a l f plane. From Equation 5.1, t h i s i s
S u b s t i t u t i n g t h i s i n t o Equation 5-2 i n d i c a t e s a maximum phase
S ince t h e maximum phase s h i f t from t h e pole-zero p a i r i n
There are two imaginary pole-zero p a i r s wi th a = b = 0. The
-1 - trrn I * -1 & a a 6 = lim [ t a n a,& 0 (5-6)
This i s always zero except f o r f requencies between t h e po le and zero ,
where 8 = 2 180'.
The a b s o l u t e s t a b i l i t y of a system designed by t h e proposed method
can now be considered without t h e u n r e a l i s t i c assumption t h a t t h e zeros of
H From t h e n a t u r e of t h e
i d e a l open loop t r a n s f e r func t ion , G ( s ) H
3-27 and 3-28, i t is apparent t h a t n o t only i s t h e Popov s t a b i l i t y c r i t e -
r i o n of Equations 2-7 o r 2-8 s a t f s f i e d for any non l inea r ga in of t he type
be ing considered, bu t it w i l l cont inue t o be s a t i s f i e d f o r any change i n
G ( s ) H ( 6 ) t h a t r e s u l t s i n a change i n the phase s h i f t of G ( s ) H (s) of eq eq
(s) can be made e x a c t l y equal t o po le s of G ( s ) . eq
(s) , as given by Equations eq
104
I n fact, t h e Popov c r i t e r i o n f o r non l inea r systems can less than 90'.
be s a t i s f i e d i n some cases f o r maximum changes g r e a t e r than 90' i f t h i s
maximum change does no t occur a t the.same frequency as t h e maximum phase
s h i f t i n t h e i d e a l case, or i f t h e s im of t h e change is such as t o re-
duce t h e t o t a l phase s h i f t .
Figures 5-2a through 5-2e i l l u s t r a t e f u r t h e r t he e f f e c t of d i f -
f e r ences i n t h e l o c a t i o n s of the p o l e and zero of a pole-zero pair. A
t h i r d o r d e r system w i t h one i n t e g r a t i o n i s used i n these examples. The
pole-zero p l o t is shown on t h e l e f t and t h e correspondinR modified f r e -
quency p l o t on t h e r i g h t . Figures 5-2d and 5-2e i n d i c a t e t h a t t h e rela-
t ive displacement of t he po le and ze ro i s important f o r pole-zero pairs
near the imaginary axis . Figure 5-2d a l s o i n d i c a t e s t he p o s s i b i l f t y of
designing f o r abso lu t e s t a b i l i t y i n a f i n i t e s e c t o r where it is n o t
p o s s i b l e t o inc lude the e n t i r e first and t h i r d quadrants i n t h e s t a b i l i t y
s e c t o r .
Systems w i t h time-varying gains must s a t i s fy s t r o n g e r cond i t ions
f o r s t a b i l i t y than those f o r non l inea r gains. Equation 2-17 i n d i c a t e s
t h a t t h e modified p l o t must never c r o s s i n t o t h e second o r t h i r d quad-
r a n t s i f t h e s t a b i l i t y s e c t o r of t h e system i s t o he i n f i n i t e . Figures
5-2b and 5-2c suggest two possible approaches i n t h e time-varvinF case*
1) I n t e n t i o n a l l y d i s p l a c e t h e zeros of H (9) as shown i n Figure 5-2b.
2) Design f o r a f i n i t e s t a b i l i t y s e c t o r . I n most cases, i n f i n i t e
s t a b i l i t y s e c t o r s are n o t required.
eq
The above d i scuss ion leads t o t h e fol lowing conclusions concerning
t h e a b s o l u t e s t a b i l i t y of a physical system designed by t h e proposed method:
105
0 X
X 0
X 0
\ # I\
Figure 5-2a
Figure 5-2b
Figure 5-2c
Figure 5-2d
n
(Continued on Next Page)
106
Figure 5-2e
Figure 5-2.
of the Pole and Zero of a Pole-Zero Pair. The Pole-Zero Diagram of a
Third Order System is shown on the L e f t , and the Corresponding Modified
Frequency Plot is Shown on the Right.
Illustrations of the Effect of Differences in the Locations
10 7
I
1. Except f o r systems with pole-zero pa i r s on o r nea r t h e
imaginary axis , t he abso lu te s t a b i l i t y s e c t o r €o r systems
wi th non l inea r Rains is i n f i n i t e .
2. Where a b s o l u t e s t a b i l i t y cannot be assured f o r an i n f i n i t e
sector because of pole-zero p a i r s nea r t h e imaginary ax i8
i t might be poss ib l e t o design f o r a f i n i t e abso lu t e sta-
b i l i t y s e c t o r .
3. I n the time-varying case , i t is n e c e s s a n t o
d i s p l a c e t h e zeros of H
an i n f i n i t e absolu te s t a b i l i t y s e c t o r .
(s) i n t h e proper d i r e c t i o n t o a s s u r e eq
The c losed loop response w a s determined i n Chapter 4 by making use
of an equ iva len t eystem cons i s t ing of an ( n - l ) s t o rder l i n e a r p a r t and a f i r s t
o r d e r nonl inear and/or t ime varying p a r t i n series.
H ( 8 ) are n o t e x a c t l y equal t o the po le s of G(s), t h e non l inea r and/or
time-varying p a r t of t h i s equivalent system has t h e form shown i n Figure
5-3.
ponding po le s , and t h e quest ion of s t r u c t u r a l s t a b i l i t y again becomes
important i n determining whether t h e output of t he equiva len t svstem i s
a v a l i d approximation t o the closed loop output of t he a c t u a l system.
Because t h e non l inea r and/or time-varying ga in cannot be separa ted i n t o
a f i r s t o rde r p a r t of t h e equiva len t system, i t does not appear p o s s i b l e
t o determine genera l a n a l y t i c a l r e s u l t s which a s s u r e t h a t t h e equ iva len t
When t h e zeros of
eq
I n t h i s equiva len t eystem, t h e zeros a r e almost equal t o t h e cor res -
system of Chapter 4 is v a l i d f o r a phys ica l syetem.
known p r o p e r t i e s of t he system i n d i c a t e t h a t t h e equ iva len t system is a
good approximation t o t h e phys ica l system. C e r t a i n a n a l y t i c a l r e s u l t s
However, a l l t he
10 8
Figure 5 - 3 .
Zeros of H
are Much Smaller Than the P
The Form of the Nonlinear Portion of the System When the
(s) are not Exactly Equal to the Poles of G(s) . The 6i p'9
i'
109
can be obtained f o r s p e c i f i c systems. For example, a r e s u l t of Zames
i
I
(1966) is used t o show t h a t t h e output of t h e system i s bounded f o r
bounded i n p u t s i f t he s t a b i l i t y sector i s no t r eau i r ed t o be i n f i n i t e
and i f only t h e p r i n c i p a l case is considered.
The r e s u l t s r epor t ed bv Z a m e s g ive s u f f i c i e n t condi t ions f o r a
bounded closed loop ou tpu t when t h e i n p u t is bounded. The cond i t ions
apply t o t h e type system considered i n t h i s d i s s e r t a t i o n i f a l l t h e p o l e s
of G ( s ) axein t h e l e f t - h a l f plane, The r e s u l t of i n t e r e s t i s c a l l e d t h e
Circle Theorem and is g iven here i n t h e following! form:
I f t h e n o n l i n e a r i t y is i n s i d e a s e c t o r {K ,K2), and i f t h e f r e - quency response of G ( s ) avoids a "c r i t i ca l region" i n t h e complex p l ane , then t h e c losed loop output i s bounded f o r bounded inpu t s : i f K1 > 0 then t h e " c r i t i c a l region" if a d i s k whose c e n t e r i s halfway between t h e p o i n t s - 1 g r e a t e r than t h e d i s t a n c e between t h e s e p o i n t s .
> - 1 + 8 , where 6 > 0.
and - - , and whose diameter i s K1 K2
cond i t ion corresponding t o t h i s "cr i t ical region" - -
K2
A g raph ica l i l l u s t r a t i o n of t h e above theorem is Riven i n Figure
5-4 . I f f (a) vs. u and G ( j w ) l i e i n t h e shaded regions, then t h e closed
loop response i s bounded f o r bounded inpu t s .
Popov c r i t e r i o n i n Chapter 2, K
t o K2.
I n t h e d i scuss ion of t h e
w a s t a k e n - t o be ze ro and K corresponds 1
From t h e above theorem and preceding d i scuss ion i t fol lows t h a t
a system designed by t h e proposed procedure, i n which G ( s ) has no po le s
on t h e imaginary axis , w i l l have a bounded closed loop response f o r
bounded i n p u t s i f t h e n o n l i n e a r i t y is confined to an appropr i a t e f i n i t e
I s e c t o r . Since t h e ga ins i n p r a c t i c a l systems are n o t i n f i n i t e , t h e
110
Figure 5-4. Illustration of the Circle Theorem. If f(u) vs.
u and G ( j w ) lie in the Shaded Regions and i f the Nyquist Diagram
of G(jo) does not Encircle the Critical Disc, the Closed Loop is
Rounded.
111
s e c t o r determined from t h i s theorem wi th appropr i a t e t o l e rances placed
on the pole and zero l o c a t i o n s w i l l be a l l t h a t i s requi red i n many
cases .
The equiva len t system of F igure 5-3 w i l l now be considered. It
i s assumed t h a t none of t h e complex conjugate pole-zero p a i r s a r e near
enough t o the imaginary axis t o cause excess ive overshoot t o occur o r
t o cause the l i n e a r system t o go uns tab le f o r s m a l l changes i n t h e pole
and zero l oca t ions . From t h i s assumption and t h e f a c t t h a t t he t r a n s f e r
func t ion between x' and x' is l i n e a r , s t a b l e , and approximately equal t o
u n i t y , one i s led t o expect the response a t x i t o be very n e a r l y equal t o
t h e output , x '
t h i s is e x a c t l y t h e case i n t h e equiva len t system of Chapter A , so it
appears poss ib l e t o use t h e equiva len t system with a hiph degree of con-
f idence except i n those cases having closed loop po le s on or near t he
imaginary a x i s . The confidence i n t h i s conclusion i s increased s t i l l
n 1
of t h e nonl inear po r t ion of t he equiva len t system. R u t n'
f u r t h s r in those cases vhe re both tho input 2nd t h e non l inea r rmd!nr t i m e -
vary ing c h a r a c t e r i s t i c are r e l a t i v e l y smooth. The requirement t h a t t h e
c losed loop poles be constrained t o be away from t h e imaginary axis i s i n
agreement wi th t h e r e s u l t s of the d i scuss ion on abso lu te s t a b i l i t y .
The f i n a l conclusion from t h e above d i scuss ion i s t h a t systems
designed by t h e proposed method can be expected t o be s t r u c t u r a l l y s t a b l e
except i n those cases having closed loop poles on o r near t h e imaginary
ax i s . Ana ly t i ca l r e s u l t s are given which i n d i c a t e t h e e x t e n t of t he e f f e c t
of s m a l l changes i n t h e pole and zero l o c a t i o n s on t h e abso lu te s t a h i l i t y
of t h e system. General a n a l y t i c a l r e s u l t s were no t obtained t o i n d i c a t e
1 1 2
t h e e x t e n t of t h e e f f e c t of such changes i n t h e case of t he closed loop
response, but reasons a r e given f o r expect ing t h e equiva len t system
developed i n Chapter 4 t o give a good approximation t o the a c t u a l c losed
loop response. The r e s u l t s of analoR computer s imula t ions , one of which
is repor t ed i n Example 4-3, support t h i s conclusion.
E f f e c t of t he Location of t he Nonlinear and/or “me-Varying Gain
The b a s i c design procedure of feeding back a l l t h e s t a t e v a r i a b l e s
i n order t o realize a des i r ed H
and/or time-varying gain i s located a r b i t r a r i l y i n t h e system.
for t h i s is i l l u s t r a t e d by PiRure 5-5a. I n t h i s system an equiva len t
feedback func t ion H ( 8 )
independent ly of t h e Rain N. No Reneral procedure f o r handl ing such Droh-
l e m s appears t o be b e s t f o r a l l systems. Consequently, it i s t r e a t e d he re
by sugges t ing the fol lowing poss ib le procedures based on t h e systems shown
i n Figures 5-5a through 5-5c.
Method 1:
can be measured and fed back, the equiva len t feedback func t ion i s
( 8 ) is not a p p l i c a b l e when the non l inea r eq
The reason
from t h e output t o t h e inpu t cannot he obtained eq
I f t he ga in N is located as shown i n Figure 5-5b and i t s output
Heq(s) = k4(s+b)(s+c)(s+d) + kg (s+c)(s+d) + k2 (s+d) + 1,
and t h e r e s u l t is t h e same as when t h e gain is loca ted as shown i n F iau re
4-1.
Method 2: I f t he ga in N i s a s a t u r a t i o n tvpe, i t i s i n some cases p o s s i b l e
t o in t roduce another s a t u r a t i o n type gain a t t h e inpu t which w i l l p revent
t h e o r i g i n a l n o n l i n e a r i t y from s a t u r a t i n g f o r any inpu t .
then equiva len t t o a system with one n o n l i n e a r i t v a t t he Droner l o c a t i o n
f o r applying the proposed design procedure.
Therefore t h e b a s i c procedure a p p l i e s .
The system i s
This method could a l s o be
t
113
. 2 1 f(0) = x 4 1
s + b s + c A N
A .
Figure 5-5a.
Gain, N. (G(s)
* 1
s f d z.-, _xl
is not Independent of
I 1
1 x2 N -s-i-c -
the Nonlinear
1 x1 m-
I I
Figure 5-5b. System Configuration for Metkod 1.
Figure 5-Sc. System Configuration for Method 2.
114
used i n some cases t o extend the desip;n procedure t o spstems with mult i -
p l e s a t u r a t i o n type n o n l i n e a r i t i e s . This approach could not be used
when pure i n t e g r a t i o n s are present between t h e input and t h e non l inea r i ty .
Method 3: For a feedback conf igura t ion such as t h a t shown i n Figure 5-5c,
t h e equiva len t feedback funct ion is
Heq(s) = k4(s+a)(s+b)(s+d) + k3(s+a)(s+d) + k2(s+d) + 1.
This i s t h e same type of equation as t h a t obtained from t h e b a s i c feedback
conf igu ra t ion of Figure 4-1, so t h a t t h e ki can be found i n t h e same manner
as before . From Figure 5-5c and t h e equat ion above, i t is apparent t h a t
H ( 8 ) cannot have zeros equal t o t h e (n-1) poles of G(s) t h a t are ou t s ide
t h e i n n e r loop with k4 as t h e feedback element, as t h i s would r equ i r e t h a t
k4 = QO.
t o feed back i n t o t h e required p o i n t s i n t h e system.
eq
The ob jec t ion t o t h i s method is t h a t it i s not u sua l lv poss ib l e
T Also, s i n c e u f-k -,-. x,
t h e matrix formulat ion of t he bastc procedure cannot be used here .
advantage is t h a t i t allows the extension of t h e b a s i c procedure t o systems
An
where series compensation i s des i r ab le and must be placed i n f r o n t of t h e
nonl inear and/or time-varying gain. I n t h i s case i t w i l l o f t e n be poss ib l e
to feed t h e output back' t o any po in t i n the- compensation network.
An equiva len t system for input-output r e l a t i o n s s i m i l a r t o t h a t
developed above can be der ived f o r a l l these methods.
from t h e previous equ iva len t system is t h a t t h e s inRle order nonl inear and/
o r time-varying p a r t w i l l appear i n t h e middle of t h e l i n e a r p a r t f o r
The only d i f f e r e n c e
Methods 1 and 3.
The determinat ion of the feedback c o e f f i c i e n t s i n Methods 1 and 2
i s no d i f f e r e n t from the bas i c procedure,
has been determined from the block diagram i n Method 3, t h e c a l c u l a t i o n
Once t h e expression f o r H (s) e u
I
of t h e k i s accomplished as in
i n t e r e s t i n g t o no te t h a t in t he
i
11s t h e b a s i c design procedure. T t is
b a s i c procedure information concerning
a l l t h e state v a r i a b l e s is fed back t o t h e inpu t i n order t o e s t a b l i s h
t h e a c t u a t i n g simal, whi le i n Method 3 t h e output is fed back so as t o
e s t a b l i s h des i r ed c o n t r o l s igna l s a t c e r t a i n p o i n t s i n t h e system.
I n many a c t u a l design problems where t h e b a s i c procedure cannot
be used, a combination of t h e above methods might be usefu l .
i n g example i l l u s t r a t e s t he procedure of Yethod 3.
Example 5-1: The f ixed p l m t of F igure 5-6a is used t o i l l u s t r a t e t h e
The follaw-
procedure of Method 3. It is assumed t h a t
a t s = - - + j &- - . The first s t e p i n the
f ixed p l a n t as i nd ica t ed i n PiRure 5-6b so
2 - 2
closed loop po les are des i r ed
des ign is t o compensate t h e
t h a t t h e open loop po le s w i l l
be a t t h e des i r ed l o c a t i o n of the c losed loop poles. The r eau i r ed feed-
back conf igu ra t ion is shown i n Fiqure 5-6c. Comparing t h i s wi th t h e
equ iva len t system of Figure 5-6d g ives
H is) - k3(S2 + s) + k2Stl 0
eq
The d e s i r e d t r a n s f e r func t ion must have t h e form
(5-71
H (s) - k3[s2 + s+l] ( 5 - 8 ) eq
Equating t h e corresponding c o e f f i c i e n t s of Equations 5-7 and 5-8 and
so lv ing t h e r e s u l t i n g equat ions g ives
k3 = 1
k2 -1
This set of feedback c o e f f i c i e n t s produces t h e d e s i r e d c losed loop po le s
116
1 - S
f - 1 N s + 2
1 s + l
A
Figure 5-6s. The Plant t o be Controlled i n Example 5-1 .
Figure 5-6b. Compensation of the Plant of Example 5-1 t o Force the Open
Loop Poles to be a t the Desired Location of the Closed Loop Poles.
4
Y 1 s + 2 - 1 -
8
-
I I
i .
Figure 5 - 6 c . Method of Controlling the Plant of Example 5-1.
1 = X
Figure 5-6d. An Equivalent System for Figure 5-6c.
117
a t s = - y + j 7 , independent of t h e nonl inear and/or time-varylng
gain.
One poin t of i n t e r e s t i s t h a t t h e feedback shown i n FiRure 5-6b \
changes the type sf the system. I f zero s teady s ta te e r r o r i s des i r ed ,
another i n t e g r a t o r must b e added t o t h e system. This would r e q u i r e an
a d d i t i o n a l feedback path as i t increases t h e o rde r of t he system and t h e
number of s ta te v a r i a b l e s by one.
Design f o r F i n i t e Sec tors of S t a b i l i t y
A l l t h e previous d iscuss ion is based on t h e premise t h a t a l l t h e
state v a r i a b l e s can be f ed back t o t h e des i r ed p o i n t s i n t h e system t o be
compensated. The systems which r e s u l t from t h e proposed d e s i m procedure
then have i n f i n i t e s t a b i l i t y s ec to r s .
back a l l t h e state v a r i a b l e s as d e s i r e d .
It is o f t e n not poss ib l e t o feed
Also, i t i s not u sua l ly necessary
t h a t t h e s t a b i l i t y s e c t o r be i n f i n i t e . In t h i s s e c t i o n , t h e p o s s i b i l i t y
of designing f o r a f i n i t e s t a b i l i t y s e c t o r when a l l t h e state v a r i a b l e s
cannot be f ed back i s discussed. (The procedure suggested i n Chapter 3
f o r those cases where a l l t he state v a r i a b l e s cannot be fed back can be
used i n t h e nonl inear case as w e l l .
t i o n a l poss ib l e porcedures.)
The suRgestions here provide addi-
Because of t h e many d i f f e r e n t s i t u a t i o n s
t h a t can arise i n nonl inear systems, no a t t e m p t i s made t o provide a
genera l so lu t ion . In s t ead , s p e c i f i c cases of nonl inear spstems are con-
s idered . The same ideas can be appl ied t o time-varying systems, but t h e
condi t ions f o r abso lu t e s t a b i l i t y a r e more str ict .
Case I:
v a r i a b l e s cannot be fed back. I f t h e o t h e r (n-1) s ta te v a r i a b l e s are
It is assumed t h a t i n an nth order system, one of t h e state
118
fed back, i t fol lows from the d iscuss ion of Chapters 3 and 4 t h a t (n-2)
of t h e ze ros of H (9) can be loca ted as des i r ed by f eed ine back through
t h e proper cons tan t gains . I f t h e s e (11-21 zeros are made equal t o po le s
of G ( s ) , t h e open loop t r a n s f e r func t ion has t h e form
ecl
K' G(s)Heq(s) (s+a) (s+b)
o r
(S -9 )
(5-11))
Theore t i ca l ly , t h e s t a b i l i t y s e c t o r i s i n f i n i t e because t h e phase s h i f t
can never be g r e a t e r than 180 , so t he Popov l i n e can always be drawn
through t h e origin. P r a c t i c a l l y , t h e r e s u l t s of t he f i r s t p a r t of t h i s
0
chap te r i n d i c a t e t h a t t he s t a b i l i t y s e c t o r might become f i n i t e because
of t h e e f f e c t of t h e zeros of H (s) no t be ing exactly equal t o Doles of
G(s) .
eq
The equiva len t system fo r input-output r e l a t i o n s i s s imilar t o t h a t
developed i n Chapter 4 except t h a t t h e non l inea r po r t ion he re i s second
o r d e r and t h e l i n e a r po r t ion i s of o rde r (n-2). The conf igura t ion of t h i s
equ iva len t system i s shown i n Figure 5-7. It i s poss ib l e t o determine t h e
output of t h e nonlinedr por t ion of t h e equiba len t system by us ing phase-
p l ane a n a l y s i s . With thfs as t h e inpu t t o t h e l i n e a r p o r t i o n of t he
equiva len t system, the output of t h e closed loop svstem can be found as
i n Example 4-3.
For t h e case under cons idera t ion , i t i s seen t h a t a b s o l u t e sta-
b i l i t y can be assured f o r a f i n i t e s e c t o r and t h a t t h e output can be found
f o r a given input . It is obvious t h a t t he e f f e c t of t h e n o n l i n e a r i t y can-
not be c o n t r o l l e d t o the ex ten t t h a t i t can i n t h e case where a l l t he
119
Figure 5 - 7 .
a s i n Case I . G'(s) i s of Order n-2.
Equivalent System When n-1 State Variables are Fed Back
120
?
state v a r i a b l e s are fed back, because now only (11-21 of t h e c losed loop
po les can be made independent of t he non l inea r gain.
Case 11:
a b l e s cannot be fed back. I n t h i s case t h e l o c a t i o n of (n-m-1) of t h e
c losed loop poles can be made independent of t h e non l inea r gain. The
open loop t r a n s f e r func t ion w i l l have (m + 1 ) po le s and t h e non l inea r
p a r t of t h e equiva len t system w i l l be of o rde r (m + 1). Thus, a l though
a f i n i t e s e c t o r of s t a b i l i t y can be r e a l i z e d and though t h e o rde r of t h e
non l inea r po r t ion of t h e equiva len t system i s lower than t h e o rde r of
t h e a c t u a l system by (n-m-l), many of t he advantaRes of t h e des ign
procedure are l o s t . It is no lonRer p o s s i b l e t o determine t h e output of
t h e system by t h e r e l a t i v e l y simple procedures used when t h e non l inea r
po r t ion of t h e equiva len t system w a s of order 1 o r 2 . Nei ther i s i t
p o s s i b l e t o show i n genera l t h a t t h e output i s bounded f o r a bounded in-
put . Because of t h i s , i t appears t h a t t h i s procedure might be more u s e f u l
i n t h e case of a r e g u l a t o r system.
Case 111:
state v a r i a b l e s cannot be fed back is t o r e a l i z e H
i n s e r t i n g a t r a n s f e r func t ion , H '(SI, i n the feedback pa th t o produce
the des i r ed zeros . I n o rde r t o make t h i s t r a n s f e r func t ion r e a l i z a b l e ,
po les must be added so t h a t H ' (9 ) w i l l have as many poles as zeros .
t hese po le s are added f a r enough out on t h e r e a l axis, they w i l l no t
apprec iab ly a f f e c t t h e system response except f o r high ga ins .
It is assumed t h a t i n an nth o rde r system, m of t h e s ta te v a r i -
Another p o s s i b l e approach t o be used when one o r more of t h e
(s) by a c t u a l l y eq
ea
If eq
The extreme example of t h i s case occurs when only t h e output can
be f ed back. Then H ' ( s ) must have (n-1) zeros and poles. Depending on eq
1 2 1
t h e e x t e n t t o which the e f f e c t of t h e poles of H ' ( 8 ) can be ignored i n
t h e reg ion of operat ion, t h i s method i s an approximation t o feedine back
t h e output and i t s (n-1) de r iva t ives (or t h e phase va r i ab le s ) i n a l i n e a r
combination. An advantage of the method i s t h a t t h e loca t ion of t h e f ixed
c losed loop poles w i l l be independent of ga ins loca ted anywhere i n t h e
system. The following example i s an i l l u s t r a t i o n of t h i s procedure.
Example 5-2:
w i th N represent ing a nonl inear gain.
t h e output can be fed back, and t h a t it can be fed back only t o the inpu t .
I f t h e s t a b i l i t y s e c t o r i s not requi red t o be i n f i n i t e , t h e r e a l i z a t i o n
of H ' ( 8 ) as t h e feedback t r a n s f e r func t ion can be used t o ge t s i m i l a r
r e s u l t s , w i th t h e primary d i f f e rence i n t h e two systems being t h e same as
those poin ted out i n t h e above d iscuss ion of t h i s method.
H '(s) might be chosen as
eq
Here t h e same f ixed p l a n t i s considered as i n Example 5-1,
However, i t i s assumed t h a t only
eq
For example,
eq A
625 (sz+e+l) H e p ) IC 2 O
(s+25) (5 -9)
The open loop t r a n s f e r funct ion of t h e l i n e a r p a r t of t h e system then
becomes
(5-10)
It has been shown by Brockett and W i l l e m s (1965a) t h a t t h e s t a b i l i t y
s e c t o r of a t h i r d o rde r nonl inear system wi th no zeros is t h e same as
the linear s t a b i l i t y s ec to r .
s e c t o r of t h e above system is (0,58.3].
less than t h a t i nd ica t ed by Equation 5-9, t h i s s t a b i l i t y s e c t o r can be
increased by t he same r a t i o .
From t h i s , i t fol lows t h a t t h e s t a b i l i t y
By makinR t h e gain of H ' ( 8 ) eq
122
summary
I n t h i s chap te r t h e p r o p e r t i e s of t he systems which r e s u l t from
t h e b a s i c design procedure are inves t lga t ed from a p r a c t i c a l viewpoint ,
I n p a r t i c u l a r , t h e e f f e c t of t he zeros of H
t o t h e po le s of G ( s ) i s considered.
s t r u c t u r a l l y s t a b l e i n s o f a r as t h i s e f f e c t i s concerned except i n those
cases having c l o s e d loop poles on o r near t h e imaginary axis.
(s) no t be ine e x a c t l y equal eq
It i s concluded t h a t t h e systems are
Poss ib l e procedures t h a t might be u s e f u l when t h e b a s i c d e s i m
method I s no t app l i cab le , e i t h e r because of t h e l o c a t i o n of t h e nonl inear-
ity in t h e system o r because a l l t h e s ta te v a r i a b l e s cannot be fed back,
a r e d iscussed . Although no general sol-ution t o t h i s problem i s obta ined ,
t h e r e are several approaches sugges t ed . The b e s t approach w i l l depend
upon t h e p a r t i c u l a r system t o be c o n t r o l l e d and t h e requi red performance
of t h e system.
CHAPTER VI
CONCLUSIONS AND SUGGESTIONS FOR FURTHER WORK
C onc lu s ion s
Design procedures for single input, single output systems based
on the modern control theory concept of feeding back all the state vari-
ables have been developed, and the characteristics of the resulting
systems studied.
method as developed by Schultz (1966).
of series compensation as practiced in classical control theory and
state variable feedback as suggested by modern control theory.
procedure for designing linear systems for a desired closed loop transfer
function is developed from the matrix representation of the system. From
this, a procedure for linear gain insensitive systems is developed.
This procedure for gain insensitive systems is then used to deveiop the
principal results of this dissertation, a method for designing certain
nonlinear and/or time-varying systems.
These procedures are based on the concept of the SVF
They utilize both the concept
First, a
The basic design procedure applies to systems with a single
memoryless nonlinear and/or time-varying gain whose input-output graph
is confined to the first and third quadrants.
linear systems to which the Popov stability criterion applies.
basic form, it is limited t o systems with the nonlinear and/or time-
varying gain located as shown in Figure 4-1.
systems classified a6 the principal case and the simplest particular
case.
This is the class of non-
In its
It is applicable only to
Modifications of the basic procedure which can be used to overcome
12 3
124
The requirement concerning t h e l o c a t i o n of t h e n o n l i n e a r i t y are d is -
cussed.
S i g n i f i c a n t f e a t u r e s of the b a s i c design procedure and t h e
r e s u l t i n g systems are as follows:
1. The procedure is app l i cab le , i n a p r a c t i c a l s ense , t o
systems of any order . Basic mat r ix o r block diagram
manipulat ions and the s o l u t i o n of a l i n e a r a l g e b r a i c
equat ions are required i n ca r ry ing i t ou t .
2. The s e c t o r of absolu te s t a b i l i t y i s i n f i n i t e .
3.
4.
The output of t h e system i s bounded f o r bounded inpu t s .
The c losed loop system can be represented by an equiva len t
system f o r studying input-output r e l a t i o n s which c o n s i s t s
of a f i r s t o rde r nonl inear and/or time-varying p o r t i o n
I n series wi th an (n-1)s t o rder s t a b l e l i n e a r por t ion . This
equiva len t system can be used t o determine t h e c losed loop
response for knom inpu t s .
5. The c losed loop system i s l i n e a r i z e d t o the e x t e n t t h a t
(n-1) of t he closed loop poles a r e equal t o (n-1) of t he open
loop po le s , independent ,of t h e non l inea r gain.
6. I n systems where the ope ra t ion i s l i n e a r f o r normal
c o n t r o l s i g n a l s but non l inea r f o r l a r g e c o n t r o l s i g n a l s ,
i t is poss ib l e t o design f o r des i r ed performance i n the
l i n e a r reg ion without having t o worry about system
s t a b i l i t y f o r large c o n t r o l simal-s.
125
Suggestions for Further Work
Since the proposed design procedure appears to have immediate
practical applications, one obvious area of further work is in the
application of the method to actual control problems.
predict what type of problems might be encountered until this is done.
It is difficult to
There are several theoretical questions concerning the method
which need to be investigated. The sensitivity of the system response
to parameter changes in the linear plant need to be investigated. Morgan
(1963, 1966) has studied this problem, and Bob White, a graduate student
at the University of Arizona, is currently investigating the subject of
sensitivity in systems with state variable feedback.
The equivalent system allows the calculation of only the output
variable and the control signal, as the other state variables in this
equivalent system are not physical variables.
magnitude of these state variables will in general probably not be
excessive compared to corresponding magnitudes in a system with unity
feedback from the output only. However, this is largely supposition,
and the subject needs to be investigated.
linear element can be found from the equivalent system, this could be
used in the actual system block diagram to determine the value of any
desired state variable.
Indications are that the
Since the output of the non-
There needs to be more of a comparison between the performance
of systems designed by the proposed method and those designed by
classical methods (or by other modem control methods).
for this comparison must be chosen.
A criterion
This problem will be resolved at
126
least in part by the success with which the procedure can be applied to
actual systems.
The modifications to the basic procedure suggested in Chapter 5
need to be investigated more thoroughly. For example, if all the state
variables cannot be fed back, is the method of determining feedback
transfer functions suggested in Chapter 3 or that of realizing an
H ' (e) as suggested in Chapter 5 better for a particular application?
More information on haw to choose a desired closed loop transfer
Likewise in the nonlinear
eq
function in the linear case would be helpful.
case, more information on how to choose the location of the closed loop
poles that are made independent of the nonlinear gain would be helpful.
It is shown that the linear gain insensitive systems always
satisfy Kalman's frequency condition for some performance index.
be interesting, and perhaps useful, to know something about the performance
index for which this condition is satisfied and whether it can be related
to the closed loop transfer function.
It would
More information concerning the extent of the effect of small
changes in the pole and zero locations on the validity of the closed loop
response as calculated from the equivalent system is needed. Additional
analog computer simulations would be helpful at this point and would
also provide other useful information concerning systems designed by the
proposed method.
In conclusion, although several aspects of systems designed by
the proposed method still need to be investigated, it is felt that the
analysis of such systems presented in this dissertation is enough to
indicate the possible usefulness of this approach to the design of control
systems with nonlinear and/or time-varying gains.
a
LIST OF REFERENCES
Aizerman, M. A. and F. R. Gantmacher. Absolute Stability of Regulator Systems. San Francisco: Holden-Day, Inc., 1964.
Bongiorno, J. J., Jr. "Real-Frequency Stability Criteria for Linear Time-Varying Systems," Proc. of the IEEE, Vol. 52, No. 7, July, 1964, 832-841.
Bower, J. L. and P. M. Schultheiss. Introduction to the Design of Servomechanisms. New York: John Wiley & Sons, Inc., 1958.
Brockett, R. W. "Poles, Zeros, and Feedback: State Spaee Interpretation," IEEE Transactions on Automatic Control, Vol. AC-10, No. 2, April, 1965, 129-135.
Brockett, R. W. and J. L. Willems. "Frequency Domain Stability Criteria, Part I," IEEE Transactions on Automatic Control, Vol. AC-10, No. 3 , July, 1965a, 255-261
Brockett, R. W. and J. L. Willems. "Frequency Domain Stability Criteria, Part 11," IEEE Transactions on Automatic Control, Vol. AC-10, No. 4, October, 1965b, 407-413.
Cunningham, W. J. Introduction to Nonlinear Analysis. New York: McGraw-Hill Book Co., Inc., 1958.
D'AZZO, J. J. and C. H. Houpis. Feedback Control System Analysis and Synthesis. New York: McGraw-Hill Book Co., Inc., 1960.
Dewey, A . G. "On the Stability of Feedback Systems with One Differen- tiable Nonlinear Element," IEEE Transactions on Automatic Control, V O ~ . AC-11, NO. 3, July, 1966, 485-491.
Dewey, A. G. and E. I. Jury. "A Stability Inequality for a Class of Nonlinear Feedback Systems," IEEE Transactions on Automatic Control, Vol. AC-11, No. 1, January, 1966, 54-62.
Gantmacher, F. R. The Theory of Matrices, Vol. 1. New York: Chelsea, 1959.
Gibson, J. E. Nonlinear Automatic Control. New York: McGraw-Hill Book Co., Inc., 1963.
127
Hahn, Wolfgang. Theory and Applications of Liapunov's Direct Nethod. Englewood Cliffs, N. J . : Prentice Hall, Inc . , 1963.
Higgins, W. T., Jr. The Stability of Certain Nonlinear, Time-Varyin3 Systems of Automatic Control. Tucson, Arizona: University of Arizona Ph.D. Dissertation, January, 1966.
Kalman, R. E. "Liapunov Functions for the Problem of Lur'e in Auto- matic Control," Proc. of the National Academy of Sciences, Vol. 49, No. 2, February, 1963a, 201-205.
Kalman, R. E. ''Mathematical Description of Linear Dynamical Systems," SIAM Journal of Control, Ser. A, Vol. 1, No. 2, 196313, 152-192.
Kalman, R. E. "When is a Linear System Optimal?" 3 . of Basic Engr., Series D, Vol. 86, March, 1964, 51-60.
Kalman, R. E. and J. E . Bertram. "Control System Analysis and Design Via the 'Second Method' of Liapunov, Part I, Continuous-Time Systems,'' Transactions of ASME, J. of Basic Engr., June, 1960, 371-393.
Kreindler, E. and P. E. Sarachick. "On the Concepts of Controllability and Observability of Linear Systems," IEEE Transactions of Automatic Control, Vol. AC-9, No. 1, January, 1964, 129-136.
LaSalle, J. P. and S. Lefschetz. Stability by Liapunov's Direct Method with Applications. New York: Academic Press, 1961.
Lefschetz, Solomon. Stability of Nonlinear Control Systems, New York: Academic Press, 1965.
Letov, A. M. Stability in Nonlinear Control Systems. Princeton, N. J.: Princeton University Press, 1961.
Margolis, S. G. and W. G. Vogt. "Control Engineering Applications of V. I. Zubov's Construction Procedure foz Liapunov Functions,'' IEEE Transactions on Automatic Control, Vol. AC-8, No, 2, April , 1963, 104-113.
Morgan, B. S., Jr, "The Synthesis of Single Variable Systems by State Variable Feedback," Proceedings of Allerton Conference on Circuit and System Theory, University of Illinois, Urbana, 1963.
Morgan, B. S., Jr. "Sensitivity Analysis and Synthesis of Multivariable Systems," IEEE Transactions on Automatic Control, Vol. AC-11, NO. 3, July, 1966, 506-512.
t
129
i
1
Narendra, K. S . and R. M. Goldwyn. "A Geometrical Criterion for the Stability of Certain Nonlinear Nonautonomous Systems," IEEE Transactions of the Circuit T?mory Group, Vol. CT-11, No. 3, September, 1964, 406-408.
Nering, E . D. Linear Algebra and Matrix Theory. New York: John Wiley & Sons, Inc., 1963.
Popov, V. M. "Absolute Stability of Nonlinear Systems of Automatic Control," Automation and Remote Control, Vol. 22, No. 8, August, 1961, 857-875.
Rozenvasser, E. M. "The Absolute Stability of Nonlinear Systems,'' Automation and Remote Control, Vol. 24, No. 3, March, 1963, 283-291.
Sandberg, I. W. "A Frequency-Domain Condition for the Stability of Feedback Systems Containing a Single Time-Varying Nonlinear Element," The Bell System Technical Journal, Vol. 43, Part 2, July, 1964, 1601-1608.
Schultz,.D..G.. "Control System Design by.State Variable Feedback Techniques," Engineering Research Lab., University of Arizona, Tucson, Annual Status Report, Vol. 1, Grant NsG-490, July, 1966.
Schultz, D. G. and J. E. Gibson. "The Variable Gradient Method for Generating Liapunov Functions," Trans. AIEE, Vol. 81, Part 11, 1963, 203-210
Truxal, J. G. Automatic Feedback Control System Synthesis. New York: McGraw-Hill Book Co., Inc . , 1955.
Yakubovich, V. A. "The Solution of Certain Matrix Inequaiities in Automatic Control Theory,'' Soviet Mathematics, Vol. 3, No. 2 , March, 1962, 620-623.
, Yakubwich, V. A. "Frequency Conditions for, the Absolute Stability
and Diasipativity of Control Systems with a Single Differentiable Nonlinearity," Soviet Mathematics, Vol. 6, No. 1, Jan.-Feb., . 1965a, 98-101
Yakubovich, V. A. "The Method of Matrix Inequalties in the Stability Theory of Nonlinear Control Systems, 11. Absolute Stability in a Class of Nonlinearities with a Condition on the Derivative," Automation and Remote Control, Vol. 26, No. 4 , April, 1965b, 577-592.
Zames, G. "On the Input-Output Stability of Time-Varying Nonlinear Feedback Systems--Part 11: Conditions Involving Circles in the Frequency Plane and Sector Nonlinearities," IEEE Transactions on Automatic Control, Vol. AC-11, No. 3, July, 1966, 465-476.